| /* |
| * Copyright (C) 2012 The Android Open Source Project |
| * |
| * Licensed under the Apache License, Version 2.0 (the "License"); |
| * you may not use this file except in compliance with the License. |
| * You may obtain a copy of the License at |
| * |
| * http://www.apache.org/licenses/LICENSE-2.0 |
| * |
| * Unless required by applicable law or agreed to in writing, software |
| * distributed under the License is distributed on an "AS IS" BASIS, |
| * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| * See the License for the specific language governing permissions and |
| * limitations under the License. |
| */ |
| |
| #define LOG_TAG "VelocityTracker" |
| //#define LOG_NDEBUG 0 |
| |
| // Log debug messages about velocity tracking. |
| #define DEBUG_VELOCITY 0 |
| |
| // Log debug messages about the progress of the algorithm itself. |
| #define DEBUG_STRATEGY 0 |
| |
| #include <inttypes.h> |
| #include <limits.h> |
| #include <math.h> |
| |
| #include <android-base/stringprintf.h> |
| #include <cutils/properties.h> |
| #include <input/VelocityTracker.h> |
| #include <utils/BitSet.h> |
| #include <utils/Timers.h> |
| |
| namespace android { |
| |
| // Nanoseconds per milliseconds. |
| static const nsecs_t NANOS_PER_MS = 1000000; |
| |
| // Threshold for determining that a pointer has stopped moving. |
| // Some input devices do not send ACTION_MOVE events in the case where a pointer has |
| // stopped. We need to detect this case so that we can accurately predict the |
| // velocity after the pointer starts moving again. |
| static const nsecs_t ASSUME_POINTER_STOPPED_TIME = 40 * NANOS_PER_MS; |
| |
| |
| static float vectorDot(const float* a, const float* b, uint32_t m) { |
| float r = 0; |
| for (size_t i = 0; i < m; i++) { |
| r += *(a++) * *(b++); |
| } |
| return r; |
| } |
| |
| static float vectorNorm(const float* a, uint32_t m) { |
| float r = 0; |
| for (size_t i = 0; i < m; i++) { |
| float t = *(a++); |
| r += t * t; |
| } |
| return sqrtf(r); |
| } |
| |
| #if DEBUG_STRATEGY || DEBUG_VELOCITY |
| static std::string vectorToString(const float* a, uint32_t m) { |
| std::string str; |
| str += "["; |
| for (size_t i = 0; i < m; i++) { |
| if (i) { |
| str += ","; |
| } |
| str += android::base::StringPrintf(" %f", *(a++)); |
| } |
| str += " ]"; |
| return str; |
| } |
| #endif |
| |
| #if DEBUG_STRATEGY |
| static std::string matrixToString(const float* a, uint32_t m, uint32_t n, bool rowMajor) { |
| std::string str; |
| str = "["; |
| for (size_t i = 0; i < m; i++) { |
| if (i) { |
| str += ","; |
| } |
| str += " ["; |
| for (size_t j = 0; j < n; j++) { |
| if (j) { |
| str += ","; |
| } |
| str += android::base::StringPrintf(" %f", a[rowMajor ? i * n + j : j * m + i]); |
| } |
| str += " ]"; |
| } |
| str += " ]"; |
| return str; |
| } |
| #endif |
| |
| |
| // --- VelocityTracker --- |
| |
| // The default velocity tracker strategy. |
| // Although other strategies are available for testing and comparison purposes, |
| // this is the strategy that applications will actually use. Be very careful |
| // when adjusting the default strategy because it can dramatically affect |
| // (often in a bad way) the user experience. |
| const char* VelocityTracker::DEFAULT_STRATEGY = "lsq2"; |
| |
| VelocityTracker::VelocityTracker(const char* strategy) : |
| mLastEventTime(0), mCurrentPointerIdBits(0), mActivePointerId(-1) { |
| char value[PROPERTY_VALUE_MAX]; |
| |
| // Allow the default strategy to be overridden using a system property for debugging. |
| if (!strategy) { |
| int length = property_get("persist.input.velocitytracker.strategy", value, nullptr); |
| if (length > 0) { |
| strategy = value; |
| } else { |
| strategy = DEFAULT_STRATEGY; |
| } |
| } |
| |
| // Configure the strategy. |
| if (!configureStrategy(strategy)) { |
| ALOGD("Unrecognized velocity tracker strategy name '%s'.", strategy); |
| if (!configureStrategy(DEFAULT_STRATEGY)) { |
| LOG_ALWAYS_FATAL("Could not create the default velocity tracker strategy '%s'!", |
| strategy); |
| } |
| } |
| } |
| |
| VelocityTracker::~VelocityTracker() { |
| delete mStrategy; |
| } |
| |
| bool VelocityTracker::configureStrategy(const char* strategy) { |
| mStrategy = createStrategy(strategy); |
| return mStrategy != nullptr; |
| } |
| |
| VelocityTrackerStrategy* VelocityTracker::createStrategy(const char* strategy) { |
| if (!strcmp("impulse", strategy)) { |
| // Physical model of pushing an object. Quality: VERY GOOD. |
| // Works with duplicate coordinates, unclean finger liftoff. |
| return new ImpulseVelocityTrackerStrategy(); |
| } |
| if (!strcmp("lsq1", strategy)) { |
| // 1st order least squares. Quality: POOR. |
| // Frequently underfits the touch data especially when the finger accelerates |
| // or changes direction. Often underestimates velocity. The direction |
| // is overly influenced by historical touch points. |
| return new LeastSquaresVelocityTrackerStrategy(1); |
| } |
| if (!strcmp("lsq2", strategy)) { |
| // 2nd order least squares. Quality: VERY GOOD. |
| // Pretty much ideal, but can be confused by certain kinds of touch data, |
| // particularly if the panel has a tendency to generate delayed, |
| // duplicate or jittery touch coordinates when the finger is released. |
| return new LeastSquaresVelocityTrackerStrategy(2); |
| } |
| if (!strcmp("lsq3", strategy)) { |
| // 3rd order least squares. Quality: UNUSABLE. |
| // Frequently overfits the touch data yielding wildly divergent estimates |
| // of the velocity when the finger is released. |
| return new LeastSquaresVelocityTrackerStrategy(3); |
| } |
| if (!strcmp("wlsq2-delta", strategy)) { |
| // 2nd order weighted least squares, delta weighting. Quality: EXPERIMENTAL |
| return new LeastSquaresVelocityTrackerStrategy(2, |
| LeastSquaresVelocityTrackerStrategy::WEIGHTING_DELTA); |
| } |
| if (!strcmp("wlsq2-central", strategy)) { |
| // 2nd order weighted least squares, central weighting. Quality: EXPERIMENTAL |
| return new LeastSquaresVelocityTrackerStrategy(2, |
| LeastSquaresVelocityTrackerStrategy::WEIGHTING_CENTRAL); |
| } |
| if (!strcmp("wlsq2-recent", strategy)) { |
| // 2nd order weighted least squares, recent weighting. Quality: EXPERIMENTAL |
| return new LeastSquaresVelocityTrackerStrategy(2, |
| LeastSquaresVelocityTrackerStrategy::WEIGHTING_RECENT); |
| } |
| if (!strcmp("int1", strategy)) { |
| // 1st order integrating filter. Quality: GOOD. |
| // Not as good as 'lsq2' because it cannot estimate acceleration but it is |
| // more tolerant of errors. Like 'lsq1', this strategy tends to underestimate |
| // the velocity of a fling but this strategy tends to respond to changes in |
| // direction more quickly and accurately. |
| return new IntegratingVelocityTrackerStrategy(1); |
| } |
| if (!strcmp("int2", strategy)) { |
| // 2nd order integrating filter. Quality: EXPERIMENTAL. |
| // For comparison purposes only. Unlike 'int1' this strategy can compensate |
| // for acceleration but it typically overestimates the effect. |
| return new IntegratingVelocityTrackerStrategy(2); |
| } |
| if (!strcmp("legacy", strategy)) { |
| // Legacy velocity tracker algorithm. Quality: POOR. |
| // For comparison purposes only. This algorithm is strongly influenced by |
| // old data points, consistently underestimates velocity and takes a very long |
| // time to adjust to changes in direction. |
| return new LegacyVelocityTrackerStrategy(); |
| } |
| return nullptr; |
| } |
| |
| void VelocityTracker::clear() { |
| mCurrentPointerIdBits.clear(); |
| mActivePointerId = -1; |
| |
| mStrategy->clear(); |
| } |
| |
| void VelocityTracker::clearPointers(BitSet32 idBits) { |
| BitSet32 remainingIdBits(mCurrentPointerIdBits.value & ~idBits.value); |
| mCurrentPointerIdBits = remainingIdBits; |
| |
| if (mActivePointerId >= 0 && idBits.hasBit(mActivePointerId)) { |
| mActivePointerId = !remainingIdBits.isEmpty() ? remainingIdBits.firstMarkedBit() : -1; |
| } |
| |
| mStrategy->clearPointers(idBits); |
| } |
| |
| void VelocityTracker::addMovement(nsecs_t eventTime, BitSet32 idBits, const Position* positions) { |
| while (idBits.count() > MAX_POINTERS) { |
| idBits.clearLastMarkedBit(); |
| } |
| |
| if ((mCurrentPointerIdBits.value & idBits.value) |
| && eventTime >= mLastEventTime + ASSUME_POINTER_STOPPED_TIME) { |
| #if DEBUG_VELOCITY |
| ALOGD("VelocityTracker: stopped for %0.3f ms, clearing state.", |
| (eventTime - mLastEventTime) * 0.000001f); |
| #endif |
| // We have not received any movements for too long. Assume that all pointers |
| // have stopped. |
| mStrategy->clear(); |
| } |
| mLastEventTime = eventTime; |
| |
| mCurrentPointerIdBits = idBits; |
| if (mActivePointerId < 0 || !idBits.hasBit(mActivePointerId)) { |
| mActivePointerId = idBits.isEmpty() ? -1 : idBits.firstMarkedBit(); |
| } |
| |
| mStrategy->addMovement(eventTime, idBits, positions); |
| |
| #if DEBUG_VELOCITY |
| ALOGD("VelocityTracker: addMovement eventTime=%" PRId64 ", idBits=0x%08x, activePointerId=%d", |
| eventTime, idBits.value, mActivePointerId); |
| for (BitSet32 iterBits(idBits); !iterBits.isEmpty(); ) { |
| uint32_t id = iterBits.firstMarkedBit(); |
| uint32_t index = idBits.getIndexOfBit(id); |
| iterBits.clearBit(id); |
| Estimator estimator; |
| getEstimator(id, &estimator); |
| ALOGD(" %d: position (%0.3f, %0.3f), " |
| "estimator (degree=%d, xCoeff=%s, yCoeff=%s, confidence=%f)", |
| id, positions[index].x, positions[index].y, |
| int(estimator.degree), |
| vectorToString(estimator.xCoeff, estimator.degree + 1).c_str(), |
| vectorToString(estimator.yCoeff, estimator.degree + 1).c_str(), |
| estimator.confidence); |
| } |
| #endif |
| } |
| |
| void VelocityTracker::addMovement(const MotionEvent* event) { |
| int32_t actionMasked = event->getActionMasked(); |
| |
| switch (actionMasked) { |
| case AMOTION_EVENT_ACTION_DOWN: |
| case AMOTION_EVENT_ACTION_HOVER_ENTER: |
| // Clear all pointers on down before adding the new movement. |
| clear(); |
| break; |
| case AMOTION_EVENT_ACTION_POINTER_DOWN: { |
| // Start a new movement trace for a pointer that just went down. |
| // We do this on down instead of on up because the client may want to query the |
| // final velocity for a pointer that just went up. |
| BitSet32 downIdBits; |
| downIdBits.markBit(event->getPointerId(event->getActionIndex())); |
| clearPointers(downIdBits); |
| break; |
| } |
| case AMOTION_EVENT_ACTION_MOVE: |
| case AMOTION_EVENT_ACTION_HOVER_MOVE: |
| break; |
| default: |
| // Ignore all other actions because they do not convey any new information about |
| // pointer movement. We also want to preserve the last known velocity of the pointers. |
| // Note that ACTION_UP and ACTION_POINTER_UP always report the last known position |
| // of the pointers that went up. ACTION_POINTER_UP does include the new position of |
| // pointers that remained down but we will also receive an ACTION_MOVE with this |
| // information if any of them actually moved. Since we don't know how many pointers |
| // will be going up at once it makes sense to just wait for the following ACTION_MOVE |
| // before adding the movement. |
| return; |
| } |
| |
| size_t pointerCount = event->getPointerCount(); |
| if (pointerCount > MAX_POINTERS) { |
| pointerCount = MAX_POINTERS; |
| } |
| |
| BitSet32 idBits; |
| for (size_t i = 0; i < pointerCount; i++) { |
| idBits.markBit(event->getPointerId(i)); |
| } |
| |
| uint32_t pointerIndex[MAX_POINTERS]; |
| for (size_t i = 0; i < pointerCount; i++) { |
| pointerIndex[i] = idBits.getIndexOfBit(event->getPointerId(i)); |
| } |
| |
| nsecs_t eventTime; |
| Position positions[pointerCount]; |
| |
| size_t historySize = event->getHistorySize(); |
| for (size_t h = 0; h < historySize; h++) { |
| eventTime = event->getHistoricalEventTime(h); |
| for (size_t i = 0; i < pointerCount; i++) { |
| uint32_t index = pointerIndex[i]; |
| positions[index].x = event->getHistoricalRawX(i, h); |
| positions[index].y = event->getHistoricalRawY(i, h); |
| } |
| addMovement(eventTime, idBits, positions); |
| } |
| |
| eventTime = event->getEventTime(); |
| for (size_t i = 0; i < pointerCount; i++) { |
| uint32_t index = pointerIndex[i]; |
| positions[index].x = event->getRawX(i); |
| positions[index].y = event->getRawY(i); |
| } |
| addMovement(eventTime, idBits, positions); |
| } |
| |
| bool VelocityTracker::getVelocity(uint32_t id, float* outVx, float* outVy) const { |
| Estimator estimator; |
| if (getEstimator(id, &estimator) && estimator.degree >= 1) { |
| *outVx = estimator.xCoeff[1]; |
| *outVy = estimator.yCoeff[1]; |
| return true; |
| } |
| *outVx = 0; |
| *outVy = 0; |
| return false; |
| } |
| |
| bool VelocityTracker::getEstimator(uint32_t id, Estimator* outEstimator) const { |
| return mStrategy->getEstimator(id, outEstimator); |
| } |
| |
| |
| // --- LeastSquaresVelocityTrackerStrategy --- |
| |
| LeastSquaresVelocityTrackerStrategy::LeastSquaresVelocityTrackerStrategy( |
| uint32_t degree, Weighting weighting) : |
| mDegree(degree), mWeighting(weighting) { |
| clear(); |
| } |
| |
| LeastSquaresVelocityTrackerStrategy::~LeastSquaresVelocityTrackerStrategy() { |
| } |
| |
| void LeastSquaresVelocityTrackerStrategy::clear() { |
| mIndex = 0; |
| mMovements[0].idBits.clear(); |
| } |
| |
| void LeastSquaresVelocityTrackerStrategy::clearPointers(BitSet32 idBits) { |
| BitSet32 remainingIdBits(mMovements[mIndex].idBits.value & ~idBits.value); |
| mMovements[mIndex].idBits = remainingIdBits; |
| } |
| |
| void LeastSquaresVelocityTrackerStrategy::addMovement(nsecs_t eventTime, BitSet32 idBits, |
| const VelocityTracker::Position* positions) { |
| if (++mIndex == HISTORY_SIZE) { |
| mIndex = 0; |
| } |
| |
| Movement& movement = mMovements[mIndex]; |
| movement.eventTime = eventTime; |
| movement.idBits = idBits; |
| uint32_t count = idBits.count(); |
| for (uint32_t i = 0; i < count; i++) { |
| movement.positions[i] = positions[i]; |
| } |
| } |
| |
| /** |
| * Solves a linear least squares problem to obtain a N degree polynomial that fits |
| * the specified input data as nearly as possible. |
| * |
| * Returns true if a solution is found, false otherwise. |
| * |
| * The input consists of two vectors of data points X and Y with indices 0..m-1 |
| * along with a weight vector W of the same size. |
| * |
| * The output is a vector B with indices 0..n that describes a polynomial |
| * that fits the data, such the sum of W[i] * W[i] * abs(Y[i] - (B[0] + B[1] X[i] |
| * + B[2] X[i]^2 ... B[n] X[i]^n)) for all i between 0 and m-1 is minimized. |
| * |
| * Accordingly, the weight vector W should be initialized by the caller with the |
| * reciprocal square root of the variance of the error in each input data point. |
| * In other words, an ideal choice for W would be W[i] = 1 / var(Y[i]) = 1 / stddev(Y[i]). |
| * The weights express the relative importance of each data point. If the weights are |
| * all 1, then the data points are considered to be of equal importance when fitting |
| * the polynomial. It is a good idea to choose weights that diminish the importance |
| * of data points that may have higher than usual error margins. |
| * |
| * Errors among data points are assumed to be independent. W is represented here |
| * as a vector although in the literature it is typically taken to be a diagonal matrix. |
| * |
| * That is to say, the function that generated the input data can be approximated |
| * by y(x) ~= B[0] + B[1] x + B[2] x^2 + ... + B[n] x^n. |
| * |
| * The coefficient of determination (R^2) is also returned to describe the goodness |
| * of fit of the model for the given data. It is a value between 0 and 1, where 1 |
| * indicates perfect correspondence. |
| * |
| * This function first expands the X vector to a m by n matrix A such that |
| * A[i][0] = 1, A[i][1] = X[i], A[i][2] = X[i]^2, ..., A[i][n] = X[i]^n, then |
| * multiplies it by w[i]./ |
| * |
| * Then it calculates the QR decomposition of A yielding an m by m orthonormal matrix Q |
| * and an m by n upper triangular matrix R. Because R is upper triangular (lower |
| * part is all zeroes), we can simplify the decomposition into an m by n matrix |
| * Q1 and a n by n matrix R1 such that A = Q1 R1. |
| * |
| * Finally we solve the system of linear equations given by R1 B = (Qtranspose W Y) |
| * to find B. |
| * |
| * For efficiency, we lay out A and Q column-wise in memory because we frequently |
| * operate on the column vectors. Conversely, we lay out R row-wise. |
| * |
| * http://en.wikipedia.org/wiki/Numerical_methods_for_linear_least_squares |
| * http://en.wikipedia.org/wiki/Gram-Schmidt |
| */ |
| static bool solveLeastSquares(const float* x, const float* y, |
| const float* w, uint32_t m, uint32_t n, float* outB, float* outDet) { |
| #if DEBUG_STRATEGY |
| ALOGD("solveLeastSquares: m=%d, n=%d, x=%s, y=%s, w=%s", int(m), int(n), |
| vectorToString(x, m).c_str(), vectorToString(y, m).c_str(), |
| vectorToString(w, m).c_str()); |
| #endif |
| |
| // Expand the X vector to a matrix A, pre-multiplied by the weights. |
| float a[n][m]; // column-major order |
| for (uint32_t h = 0; h < m; h++) { |
| a[0][h] = w[h]; |
| for (uint32_t i = 1; i < n; i++) { |
| a[i][h] = a[i - 1][h] * x[h]; |
| } |
| } |
| #if DEBUG_STRATEGY |
| ALOGD(" - a=%s", matrixToString(&a[0][0], m, n, false /*rowMajor*/).c_str()); |
| #endif |
| |
| // Apply the Gram-Schmidt process to A to obtain its QR decomposition. |
| float q[n][m]; // orthonormal basis, column-major order |
| float r[n][n]; // upper triangular matrix, row-major order |
| for (uint32_t j = 0; j < n; j++) { |
| for (uint32_t h = 0; h < m; h++) { |
| q[j][h] = a[j][h]; |
| } |
| for (uint32_t i = 0; i < j; i++) { |
| float dot = vectorDot(&q[j][0], &q[i][0], m); |
| for (uint32_t h = 0; h < m; h++) { |
| q[j][h] -= dot * q[i][h]; |
| } |
| } |
| |
| float norm = vectorNorm(&q[j][0], m); |
| if (norm < 0.000001f) { |
| // vectors are linearly dependent or zero so no solution |
| #if DEBUG_STRATEGY |
| ALOGD(" - no solution, norm=%f", norm); |
| #endif |
| return false; |
| } |
| |
| float invNorm = 1.0f / norm; |
| for (uint32_t h = 0; h < m; h++) { |
| q[j][h] *= invNorm; |
| } |
| for (uint32_t i = 0; i < n; i++) { |
| r[j][i] = i < j ? 0 : vectorDot(&q[j][0], &a[i][0], m); |
| } |
| } |
| #if DEBUG_STRATEGY |
| ALOGD(" - q=%s", matrixToString(&q[0][0], m, n, false /*rowMajor*/).c_str()); |
| ALOGD(" - r=%s", matrixToString(&r[0][0], n, n, true /*rowMajor*/).c_str()); |
| |
| // calculate QR, if we factored A correctly then QR should equal A |
| float qr[n][m]; |
| for (uint32_t h = 0; h < m; h++) { |
| for (uint32_t i = 0; i < n; i++) { |
| qr[i][h] = 0; |
| for (uint32_t j = 0; j < n; j++) { |
| qr[i][h] += q[j][h] * r[j][i]; |
| } |
| } |
| } |
| ALOGD(" - qr=%s", matrixToString(&qr[0][0], m, n, false /*rowMajor*/).c_str()); |
| #endif |
| |
| // Solve R B = Qt W Y to find B. This is easy because R is upper triangular. |
| // We just work from bottom-right to top-left calculating B's coefficients. |
| float wy[m]; |
| for (uint32_t h = 0; h < m; h++) { |
| wy[h] = y[h] * w[h]; |
| } |
| for (uint32_t i = n; i != 0; ) { |
| i--; |
| outB[i] = vectorDot(&q[i][0], wy, m); |
| for (uint32_t j = n - 1; j > i; j--) { |
| outB[i] -= r[i][j] * outB[j]; |
| } |
| outB[i] /= r[i][i]; |
| } |
| #if DEBUG_STRATEGY |
| ALOGD(" - b=%s", vectorToString(outB, n).c_str()); |
| #endif |
| |
| // Calculate the coefficient of determination as 1 - (SSerr / SStot) where |
| // SSerr is the residual sum of squares (variance of the error), |
| // and SStot is the total sum of squares (variance of the data) where each |
| // has been weighted. |
| float ymean = 0; |
| for (uint32_t h = 0; h < m; h++) { |
| ymean += y[h]; |
| } |
| ymean /= m; |
| |
| float sserr = 0; |
| float sstot = 0; |
| for (uint32_t h = 0; h < m; h++) { |
| float err = y[h] - outB[0]; |
| float term = 1; |
| for (uint32_t i = 1; i < n; i++) { |
| term *= x[h]; |
| err -= term * outB[i]; |
| } |
| sserr += w[h] * w[h] * err * err; |
| float var = y[h] - ymean; |
| sstot += w[h] * w[h] * var * var; |
| } |
| *outDet = sstot > 0.000001f ? 1.0f - (sserr / sstot) : 1; |
| #if DEBUG_STRATEGY |
| ALOGD(" - sserr=%f", sserr); |
| ALOGD(" - sstot=%f", sstot); |
| ALOGD(" - det=%f", *outDet); |
| #endif |
| return true; |
| } |
| |
| /* |
| * Optimized unweighted second-order least squares fit. About 2x speed improvement compared to |
| * the default implementation |
| */ |
| static float solveUnweightedLeastSquaresDeg2(const float* x, const float* y, size_t count) { |
| float sxi = 0, sxiyi = 0, syi = 0, sxi2 = 0, sxi3 = 0, sxi2yi = 0, sxi4 = 0; |
| |
| for (size_t i = 0; i < count; i++) { |
| float xi = x[i]; |
| float yi = y[i]; |
| float xi2 = xi*xi; |
| float xi3 = xi2*xi; |
| float xi4 = xi3*xi; |
| float xi2yi = xi2*yi; |
| float xiyi = xi*yi; |
| |
| sxi += xi; |
| sxi2 += xi2; |
| sxiyi += xiyi; |
| sxi2yi += xi2yi; |
| syi += yi; |
| sxi3 += xi3; |
| sxi4 += xi4; |
| } |
| |
| float Sxx = sxi2 - sxi*sxi / count; |
| float Sxy = sxiyi - sxi*syi / count; |
| float Sxx2 = sxi3 - sxi*sxi2 / count; |
| float Sx2y = sxi2yi - sxi2*syi / count; |
| float Sx2x2 = sxi4 - sxi2*sxi2 / count; |
| |
| float numerator = Sxy*Sx2x2 - Sx2y*Sxx2; |
| float denominator = Sxx*Sx2x2 - Sxx2*Sxx2; |
| if (denominator == 0) { |
| ALOGW("division by 0 when computing velocity, Sxx=%f, Sx2x2=%f, Sxx2=%f", Sxx, Sx2x2, Sxx2); |
| return 0; |
| } |
| return numerator/denominator; |
| } |
| |
| bool LeastSquaresVelocityTrackerStrategy::getEstimator(uint32_t id, |
| VelocityTracker::Estimator* outEstimator) const { |
| outEstimator->clear(); |
| |
| // Iterate over movement samples in reverse time order and collect samples. |
| float x[HISTORY_SIZE]; |
| float y[HISTORY_SIZE]; |
| float w[HISTORY_SIZE]; |
| float time[HISTORY_SIZE]; |
| uint32_t m = 0; |
| uint32_t index = mIndex; |
| const Movement& newestMovement = mMovements[mIndex]; |
| do { |
| const Movement& movement = mMovements[index]; |
| if (!movement.idBits.hasBit(id)) { |
| break; |
| } |
| |
| nsecs_t age = newestMovement.eventTime - movement.eventTime; |
| if (age > HORIZON) { |
| break; |
| } |
| |
| const VelocityTracker::Position& position = movement.getPosition(id); |
| x[m] = position.x; |
| y[m] = position.y; |
| w[m] = chooseWeight(index); |
| time[m] = -age * 0.000000001f; |
| index = (index == 0 ? HISTORY_SIZE : index) - 1; |
| } while (++m < HISTORY_SIZE); |
| |
| if (m == 0) { |
| return false; // no data |
| } |
| |
| // Calculate a least squares polynomial fit. |
| uint32_t degree = mDegree; |
| if (degree > m - 1) { |
| degree = m - 1; |
| } |
| if (degree >= 1) { |
| if (degree == 2 && mWeighting == WEIGHTING_NONE) { // optimize unweighted, degree=2 fit |
| outEstimator->time = newestMovement.eventTime; |
| outEstimator->degree = 2; |
| outEstimator->confidence = 1; |
| outEstimator->xCoeff[0] = 0; // only slope is calculated, set rest of coefficients = 0 |
| outEstimator->yCoeff[0] = 0; |
| outEstimator->xCoeff[1] = solveUnweightedLeastSquaresDeg2(time, x, m); |
| outEstimator->yCoeff[1] = solveUnweightedLeastSquaresDeg2(time, y, m); |
| outEstimator->xCoeff[2] = 0; |
| outEstimator->yCoeff[2] = 0; |
| return true; |
| } |
| |
| float xdet, ydet; |
| uint32_t n = degree + 1; |
| if (solveLeastSquares(time, x, w, m, n, outEstimator->xCoeff, &xdet) |
| && solveLeastSquares(time, y, w, m, n, outEstimator->yCoeff, &ydet)) { |
| outEstimator->time = newestMovement.eventTime; |
| outEstimator->degree = degree; |
| outEstimator->confidence = xdet * ydet; |
| #if DEBUG_STRATEGY |
| ALOGD("estimate: degree=%d, xCoeff=%s, yCoeff=%s, confidence=%f", |
| int(outEstimator->degree), |
| vectorToString(outEstimator->xCoeff, n).c_str(), |
| vectorToString(outEstimator->yCoeff, n).c_str(), |
| outEstimator->confidence); |
| #endif |
| return true; |
| } |
| } |
| |
| // No velocity data available for this pointer, but we do have its current position. |
| outEstimator->xCoeff[0] = x[0]; |
| outEstimator->yCoeff[0] = y[0]; |
| outEstimator->time = newestMovement.eventTime; |
| outEstimator->degree = 0; |
| outEstimator->confidence = 1; |
| return true; |
| } |
| |
| float LeastSquaresVelocityTrackerStrategy::chooseWeight(uint32_t index) const { |
| switch (mWeighting) { |
| case WEIGHTING_DELTA: { |
| // Weight points based on how much time elapsed between them and the next |
| // point so that points that "cover" a shorter time span are weighed less. |
| // delta 0ms: 0.5 |
| // delta 10ms: 1.0 |
| if (index == mIndex) { |
| return 1.0f; |
| } |
| uint32_t nextIndex = (index + 1) % HISTORY_SIZE; |
| float deltaMillis = (mMovements[nextIndex].eventTime- mMovements[index].eventTime) |
| * 0.000001f; |
| if (deltaMillis < 0) { |
| return 0.5f; |
| } |
| if (deltaMillis < 10) { |
| return 0.5f + deltaMillis * 0.05; |
| } |
| return 1.0f; |
| } |
| |
| case WEIGHTING_CENTRAL: { |
| // Weight points based on their age, weighing very recent and very old points less. |
| // age 0ms: 0.5 |
| // age 10ms: 1.0 |
| // age 50ms: 1.0 |
| // age 60ms: 0.5 |
| float ageMillis = (mMovements[mIndex].eventTime - mMovements[index].eventTime) |
| * 0.000001f; |
| if (ageMillis < 0) { |
| return 0.5f; |
| } |
| if (ageMillis < 10) { |
| return 0.5f + ageMillis * 0.05; |
| } |
| if (ageMillis < 50) { |
| return 1.0f; |
| } |
| if (ageMillis < 60) { |
| return 0.5f + (60 - ageMillis) * 0.05; |
| } |
| return 0.5f; |
| } |
| |
| case WEIGHTING_RECENT: { |
| // Weight points based on their age, weighing older points less. |
| // age 0ms: 1.0 |
| // age 50ms: 1.0 |
| // age 100ms: 0.5 |
| float ageMillis = (mMovements[mIndex].eventTime - mMovements[index].eventTime) |
| * 0.000001f; |
| if (ageMillis < 50) { |
| return 1.0f; |
| } |
| if (ageMillis < 100) { |
| return 0.5f + (100 - ageMillis) * 0.01f; |
| } |
| return 0.5f; |
| } |
| |
| case WEIGHTING_NONE: |
| default: |
| return 1.0f; |
| } |
| } |
| |
| |
| // --- IntegratingVelocityTrackerStrategy --- |
| |
| IntegratingVelocityTrackerStrategy::IntegratingVelocityTrackerStrategy(uint32_t degree) : |
| mDegree(degree) { |
| } |
| |
| IntegratingVelocityTrackerStrategy::~IntegratingVelocityTrackerStrategy() { |
| } |
| |
| void IntegratingVelocityTrackerStrategy::clear() { |
| mPointerIdBits.clear(); |
| } |
| |
| void IntegratingVelocityTrackerStrategy::clearPointers(BitSet32 idBits) { |
| mPointerIdBits.value &= ~idBits.value; |
| } |
| |
| void IntegratingVelocityTrackerStrategy::addMovement(nsecs_t eventTime, BitSet32 idBits, |
| const VelocityTracker::Position* positions) { |
| uint32_t index = 0; |
| for (BitSet32 iterIdBits(idBits); !iterIdBits.isEmpty();) { |
| uint32_t id = iterIdBits.clearFirstMarkedBit(); |
| State& state = mPointerState[id]; |
| const VelocityTracker::Position& position = positions[index++]; |
| if (mPointerIdBits.hasBit(id)) { |
| updateState(state, eventTime, position.x, position.y); |
| } else { |
| initState(state, eventTime, position.x, position.y); |
| } |
| } |
| |
| mPointerIdBits = idBits; |
| } |
| |
| bool IntegratingVelocityTrackerStrategy::getEstimator(uint32_t id, |
| VelocityTracker::Estimator* outEstimator) const { |
| outEstimator->clear(); |
| |
| if (mPointerIdBits.hasBit(id)) { |
| const State& state = mPointerState[id]; |
| populateEstimator(state, outEstimator); |
| return true; |
| } |
| |
| return false; |
| } |
| |
| void IntegratingVelocityTrackerStrategy::initState(State& state, |
| nsecs_t eventTime, float xpos, float ypos) const { |
| state.updateTime = eventTime; |
| state.degree = 0; |
| |
| state.xpos = xpos; |
| state.xvel = 0; |
| state.xaccel = 0; |
| state.ypos = ypos; |
| state.yvel = 0; |
| state.yaccel = 0; |
| } |
| |
| void IntegratingVelocityTrackerStrategy::updateState(State& state, |
| nsecs_t eventTime, float xpos, float ypos) const { |
| const nsecs_t MIN_TIME_DELTA = 2 * NANOS_PER_MS; |
| const float FILTER_TIME_CONSTANT = 0.010f; // 10 milliseconds |
| |
| if (eventTime <= state.updateTime + MIN_TIME_DELTA) { |
| return; |
| } |
| |
| float dt = (eventTime - state.updateTime) * 0.000000001f; |
| state.updateTime = eventTime; |
| |
| float xvel = (xpos - state.xpos) / dt; |
| float yvel = (ypos - state.ypos) / dt; |
| if (state.degree == 0) { |
| state.xvel = xvel; |
| state.yvel = yvel; |
| state.degree = 1; |
| } else { |
| float alpha = dt / (FILTER_TIME_CONSTANT + dt); |
| if (mDegree == 1) { |
| state.xvel += (xvel - state.xvel) * alpha; |
| state.yvel += (yvel - state.yvel) * alpha; |
| } else { |
| float xaccel = (xvel - state.xvel) / dt; |
| float yaccel = (yvel - state.yvel) / dt; |
| if (state.degree == 1) { |
| state.xaccel = xaccel; |
| state.yaccel = yaccel; |
| state.degree = 2; |
| } else { |
| state.xaccel += (xaccel - state.xaccel) * alpha; |
| state.yaccel += (yaccel - state.yaccel) * alpha; |
| } |
| state.xvel += (state.xaccel * dt) * alpha; |
| state.yvel += (state.yaccel * dt) * alpha; |
| } |
| } |
| state.xpos = xpos; |
| state.ypos = ypos; |
| } |
| |
| void IntegratingVelocityTrackerStrategy::populateEstimator(const State& state, |
| VelocityTracker::Estimator* outEstimator) const { |
| outEstimator->time = state.updateTime; |
| outEstimator->confidence = 1.0f; |
| outEstimator->degree = state.degree; |
| outEstimator->xCoeff[0] = state.xpos; |
| outEstimator->xCoeff[1] = state.xvel; |
| outEstimator->xCoeff[2] = state.xaccel / 2; |
| outEstimator->yCoeff[0] = state.ypos; |
| outEstimator->yCoeff[1] = state.yvel; |
| outEstimator->yCoeff[2] = state.yaccel / 2; |
| } |
| |
| |
| // --- LegacyVelocityTrackerStrategy --- |
| |
| LegacyVelocityTrackerStrategy::LegacyVelocityTrackerStrategy() { |
| clear(); |
| } |
| |
| LegacyVelocityTrackerStrategy::~LegacyVelocityTrackerStrategy() { |
| } |
| |
| void LegacyVelocityTrackerStrategy::clear() { |
| mIndex = 0; |
| mMovements[0].idBits.clear(); |
| } |
| |
| void LegacyVelocityTrackerStrategy::clearPointers(BitSet32 idBits) { |
| BitSet32 remainingIdBits(mMovements[mIndex].idBits.value & ~idBits.value); |
| mMovements[mIndex].idBits = remainingIdBits; |
| } |
| |
| void LegacyVelocityTrackerStrategy::addMovement(nsecs_t eventTime, BitSet32 idBits, |
| const VelocityTracker::Position* positions) { |
| if (++mIndex == HISTORY_SIZE) { |
| mIndex = 0; |
| } |
| |
| Movement& movement = mMovements[mIndex]; |
| movement.eventTime = eventTime; |
| movement.idBits = idBits; |
| uint32_t count = idBits.count(); |
| for (uint32_t i = 0; i < count; i++) { |
| movement.positions[i] = positions[i]; |
| } |
| } |
| |
| bool LegacyVelocityTrackerStrategy::getEstimator(uint32_t id, |
| VelocityTracker::Estimator* outEstimator) const { |
| outEstimator->clear(); |
| |
| const Movement& newestMovement = mMovements[mIndex]; |
| if (!newestMovement.idBits.hasBit(id)) { |
| return false; // no data |
| } |
| |
| // Find the oldest sample that contains the pointer and that is not older than HORIZON. |
| nsecs_t minTime = newestMovement.eventTime - HORIZON; |
| uint32_t oldestIndex = mIndex; |
| uint32_t numTouches = 1; |
| do { |
| uint32_t nextOldestIndex = (oldestIndex == 0 ? HISTORY_SIZE : oldestIndex) - 1; |
| const Movement& nextOldestMovement = mMovements[nextOldestIndex]; |
| if (!nextOldestMovement.idBits.hasBit(id) |
| || nextOldestMovement.eventTime < minTime) { |
| break; |
| } |
| oldestIndex = nextOldestIndex; |
| } while (++numTouches < HISTORY_SIZE); |
| |
| // Calculate an exponentially weighted moving average of the velocity estimate |
| // at different points in time measured relative to the oldest sample. |
| // This is essentially an IIR filter. Newer samples are weighted more heavily |
| // than older samples. Samples at equal time points are weighted more or less |
| // equally. |
| // |
| // One tricky problem is that the sample data may be poorly conditioned. |
| // Sometimes samples arrive very close together in time which can cause us to |
| // overestimate the velocity at that time point. Most samples might be measured |
| // 16ms apart but some consecutive samples could be only 0.5sm apart because |
| // the hardware or driver reports them irregularly or in bursts. |
| float accumVx = 0; |
| float accumVy = 0; |
| uint32_t index = oldestIndex; |
| uint32_t samplesUsed = 0; |
| const Movement& oldestMovement = mMovements[oldestIndex]; |
| const VelocityTracker::Position& oldestPosition = oldestMovement.getPosition(id); |
| nsecs_t lastDuration = 0; |
| |
| while (numTouches-- > 1) { |
| if (++index == HISTORY_SIZE) { |
| index = 0; |
| } |
| const Movement& movement = mMovements[index]; |
| nsecs_t duration = movement.eventTime - oldestMovement.eventTime; |
| |
| // If the duration between samples is small, we may significantly overestimate |
| // the velocity. Consequently, we impose a minimum duration constraint on the |
| // samples that we include in the calculation. |
| if (duration >= MIN_DURATION) { |
| const VelocityTracker::Position& position = movement.getPosition(id); |
| float scale = 1000000000.0f / duration; // one over time delta in seconds |
| float vx = (position.x - oldestPosition.x) * scale; |
| float vy = (position.y - oldestPosition.y) * scale; |
| accumVx = (accumVx * lastDuration + vx * duration) / (duration + lastDuration); |
| accumVy = (accumVy * lastDuration + vy * duration) / (duration + lastDuration); |
| lastDuration = duration; |
| samplesUsed += 1; |
| } |
| } |
| |
| // Report velocity. |
| const VelocityTracker::Position& newestPosition = newestMovement.getPosition(id); |
| outEstimator->time = newestMovement.eventTime; |
| outEstimator->confidence = 1; |
| outEstimator->xCoeff[0] = newestPosition.x; |
| outEstimator->yCoeff[0] = newestPosition.y; |
| if (samplesUsed) { |
| outEstimator->xCoeff[1] = accumVx; |
| outEstimator->yCoeff[1] = accumVy; |
| outEstimator->degree = 1; |
| } else { |
| outEstimator->degree = 0; |
| } |
| return true; |
| } |
| |
| // --- ImpulseVelocityTrackerStrategy --- |
| |
| ImpulseVelocityTrackerStrategy::ImpulseVelocityTrackerStrategy() { |
| clear(); |
| } |
| |
| ImpulseVelocityTrackerStrategy::~ImpulseVelocityTrackerStrategy() { |
| } |
| |
| void ImpulseVelocityTrackerStrategy::clear() { |
| mIndex = 0; |
| mMovements[0].idBits.clear(); |
| } |
| |
| void ImpulseVelocityTrackerStrategy::clearPointers(BitSet32 idBits) { |
| BitSet32 remainingIdBits(mMovements[mIndex].idBits.value & ~idBits.value); |
| mMovements[mIndex].idBits = remainingIdBits; |
| } |
| |
| void ImpulseVelocityTrackerStrategy::addMovement(nsecs_t eventTime, BitSet32 idBits, |
| const VelocityTracker::Position* positions) { |
| if (++mIndex == HISTORY_SIZE) { |
| mIndex = 0; |
| } |
| |
| Movement& movement = mMovements[mIndex]; |
| movement.eventTime = eventTime; |
| movement.idBits = idBits; |
| uint32_t count = idBits.count(); |
| for (uint32_t i = 0; i < count; i++) { |
| movement.positions[i] = positions[i]; |
| } |
| } |
| |
| /** |
| * Calculate the total impulse provided to the screen and the resulting velocity. |
| * |
| * The touchscreen is modeled as a physical object. |
| * Initial condition is discussed below, but for now suppose that v(t=0) = 0 |
| * |
| * The kinetic energy of the object at the release is E=0.5*m*v^2 |
| * Then vfinal = sqrt(2E/m). The goal is to calculate E. |
| * |
| * The kinetic energy at the release is equal to the total work done on the object by the finger. |
| * The total work W is the sum of all dW along the path. |
| * |
| * dW = F*dx, where dx is the piece of path traveled. |
| * Force is change of momentum over time, F = dp/dt = m dv/dt. |
| * Then substituting: |
| * dW = m (dv/dt) * dx = m * v * dv |
| * |
| * Summing along the path, we get: |
| * W = sum(dW) = sum(m * v * dv) = m * sum(v * dv) |
| * Since the mass stays constant, the equation for final velocity is: |
| * vfinal = sqrt(2*sum(v * dv)) |
| * |
| * Here, |
| * dv : change of velocity = (v[i+1]-v[i]) |
| * dx : change of distance = (x[i+1]-x[i]) |
| * dt : change of time = (t[i+1]-t[i]) |
| * v : instantaneous velocity = dx/dt |
| * |
| * The final formula is: |
| * vfinal = sqrt(2) * sqrt(sum((v[i]-v[i-1])*|v[i]|)) for all i |
| * The absolute value is needed to properly account for the sign. If the velocity over a |
| * particular segment descreases, then this indicates braking, which means that negative |
| * work was done. So for two positive, but decreasing, velocities, this contribution would be |
| * negative and will cause a smaller final velocity. |
| * |
| * Initial condition |
| * There are two ways to deal with initial condition: |
| * 1) Assume that v(0) = 0, which would mean that the screen is initially at rest. |
| * This is not entirely accurate. We are only taking the past X ms of touch data, where X is |
| * currently equal to 100. However, a touch event that created a fling probably lasted for longer |
| * than that, which would mean that the user has already been interacting with the touchscreen |
| * and it has probably already been moving. |
| * 2) Assume that the touchscreen has already been moving at a certain velocity, calculate this |
| * initial velocity and the equivalent energy, and start with this initial energy. |
| * Consider an example where we have the following data, consisting of 3 points: |
| * time: t0, t1, t2 |
| * x : x0, x1, x2 |
| * v : 0 , v1, v2 |
| * Here is what will happen in each of these scenarios: |
| * 1) By directly applying the formula above with the v(0) = 0 boundary condition, we will get |
| * vfinal = sqrt(2*(|v1|*(v1-v0) + |v2|*(v2-v1))). This can be simplified since v0=0 |
| * vfinal = sqrt(2*(|v1|*v1 + |v2|*(v2-v1))) = sqrt(2*(v1^2 + |v2|*(v2 - v1))) |
| * since velocity is a real number |
| * 2) If we treat the screen as already moving, then it must already have an energy (per mass) |
| * equal to 1/2*v1^2. Then the initial energy should be 1/2*v1*2, and only the second segment |
| * will contribute to the total kinetic energy (since we can effectively consider that v0=v1). |
| * This will give the following expression for the final velocity: |
| * vfinal = sqrt(2*(1/2*v1^2 + |v2|*(v2-v1))) |
| * This analysis can be generalized to an arbitrary number of samples. |
| * |
| * |
| * Comparing the two equations above, we see that the only mathematical difference |
| * is the factor of 1/2 in front of the first velocity term. |
| * This boundary condition would allow for the "proper" calculation of the case when all of the |
| * samples are equally spaced in time and distance, which should suggest a constant velocity. |
| * |
| * Note that approach 2) is sensitive to the proper ordering of the data in time, since |
| * the boundary condition must be applied to the oldest sample to be accurate. |
| */ |
| static float kineticEnergyToVelocity(float work) { |
| static constexpr float sqrt2 = 1.41421356237; |
| return (work < 0 ? -1.0 : 1.0) * sqrtf(fabsf(work)) * sqrt2; |
| } |
| |
| static float calculateImpulseVelocity(const nsecs_t* t, const float* x, size_t count) { |
| // The input should be in reversed time order (most recent sample at index i=0) |
| // t[i] is in nanoseconds, but due to FP arithmetic, convert to seconds inside this function |
| static constexpr float SECONDS_PER_NANO = 1E-9; |
| |
| if (count < 2) { |
| return 0; // if 0 or 1 points, velocity is zero |
| } |
| if (t[1] > t[0]) { // Algorithm will still work, but not perfectly |
| ALOGE("Samples provided to calculateImpulseVelocity in the wrong order"); |
| } |
| if (count == 2) { // if 2 points, basic linear calculation |
| if (t[1] == t[0]) { |
| ALOGE("Events have identical time stamps t=%" PRId64 ", setting velocity = 0", t[0]); |
| return 0; |
| } |
| return (x[1] - x[0]) / (SECONDS_PER_NANO * (t[1] - t[0])); |
| } |
| // Guaranteed to have at least 3 points here |
| float work = 0; |
| for (size_t i = count - 1; i > 0 ; i--) { // start with the oldest sample and go forward in time |
| if (t[i] == t[i-1]) { |
| ALOGE("Events have identical time stamps t=%" PRId64 ", skipping sample", t[i]); |
| continue; |
| } |
| float vprev = kineticEnergyToVelocity(work); // v[i-1] |
| float vcurr = (x[i] - x[i-1]) / (SECONDS_PER_NANO * (t[i] - t[i-1])); // v[i] |
| work += (vcurr - vprev) * fabsf(vcurr); |
| if (i == count - 1) { |
| work *= 0.5; // initial condition, case 2) above |
| } |
| } |
| return kineticEnergyToVelocity(work); |
| } |
| |
| bool ImpulseVelocityTrackerStrategy::getEstimator(uint32_t id, |
| VelocityTracker::Estimator* outEstimator) const { |
| outEstimator->clear(); |
| |
| // Iterate over movement samples in reverse time order and collect samples. |
| float x[HISTORY_SIZE]; |
| float y[HISTORY_SIZE]; |
| nsecs_t time[HISTORY_SIZE]; |
| size_t m = 0; // number of points that will be used for fitting |
| size_t index = mIndex; |
| const Movement& newestMovement = mMovements[mIndex]; |
| do { |
| const Movement& movement = mMovements[index]; |
| if (!movement.idBits.hasBit(id)) { |
| break; |
| } |
| |
| nsecs_t age = newestMovement.eventTime - movement.eventTime; |
| if (age > HORIZON) { |
| break; |
| } |
| |
| const VelocityTracker::Position& position = movement.getPosition(id); |
| x[m] = position.x; |
| y[m] = position.y; |
| time[m] = movement.eventTime; |
| index = (index == 0 ? HISTORY_SIZE : index) - 1; |
| } while (++m < HISTORY_SIZE); |
| |
| if (m == 0) { |
| return false; // no data |
| } |
| outEstimator->xCoeff[0] = 0; |
| outEstimator->yCoeff[0] = 0; |
| outEstimator->xCoeff[1] = calculateImpulseVelocity(time, x, m); |
| outEstimator->yCoeff[1] = calculateImpulseVelocity(time, y, m); |
| outEstimator->xCoeff[2] = 0; |
| outEstimator->yCoeff[2] = 0; |
| outEstimator->time = newestMovement.eventTime; |
| outEstimator->degree = 2; // similar results to 2nd degree fit |
| outEstimator->confidence = 1; |
| #if DEBUG_STRATEGY |
| ALOGD("velocity: (%f, %f)", outEstimator->xCoeff[1], outEstimator->yCoeff[1]); |
| #endif |
| return true; |
| } |
| |
| } // namespace android |