with class
Dependencies: ISR_Mini-explorer mbed
Fork of VirtualForces by
Diff: MiniExplorerCoimbra.cpp
- Revision:
- 39:890439b495e3
- Parent:
- 38:5ed7c79fb724
diff -r 5ed7c79fb724 -r 890439b495e3 MiniExplorerCoimbra.cpp --- a/MiniExplorerCoimbra.cpp Thu Jun 15 23:17:55 2017 +0000 +++ b/MiniExplorerCoimbra.cpp Fri Jun 16 10:40:53 2017 +0000 @@ -540,4 +540,570 @@ } return angleBeetweenRobotAndTarget-theta; } -*/ \ No newline at end of file +*/ + + +/* +Lab4 + +make the robot do the same square over and over, see that at one point the errors have accumulated and it is not where the odometria think it is +solution: +(here our sensors are bad but our odometrie s okay +before each new movement we compute an estimation of where we are, compare it to what the robot think (maybe choose our estimate?) +*/ + +void MiniExplorerCoimbra::test_procedure_lab_4(){ + this->map.fill_map_with_initial(); + float xEstimatedK=this->xWorld; + float yEstimatedK=this->yWorld; + float thetaWorldEstimatedK=this->thetaWorld; + float distanceMoved; + float angleMoved; + float PreviousCovarianceOdometricPositionEstimate[3][3]; + PreviousCovarianceOdometricPositionEstimate[0][0]=0; + PreviousCovarianceOdometricPositionEstimate[0][1]=0; + PreviousCovarianceOdometricPositionEstimate[0][2]=0; + PreviousCovarianceOdometricPositionEstimate[1][0]=0; + PreviousCovarianceOdometricPositionEstimate[1][1]=0; + PreviousCovarianceOdometricPositionEstimate[1][2]=0; + PreviousCovarianceOdometricPositionEstimate[2][0]=0; + PreviousCovarianceOdometricPositionEstimate[2][1]=0; + PreviousCovarianceOdometricPositionEstimate[2][2]=0; + while(1){ + distanceMoved=50; + angleMoved=0; + this->go_straight_line(50); + wait(1); + procedure_lab_4(xEstimatedK,yEstimatedK,thetaEstimatedK,distanceMoved,angleMoved,PreviousCovarianceOdometricPositionEstimate); + pc.printf("\n\r x:%f,y:%f,theta:%f",this->xWorld,this->yWorld,this->thetaWorld); + pc.printf("\n\r xEstim:%f,yEstim:%f,thetaEstim:%f",xEstimatedK,yEstimatedK,thetaEstimatedK); + this->turn_to_target(PI/2); + distanceMoved=0; + angleMoved=PI/2; + procedure_lab_4(xEstimatedK,yEstimatedK,thetaEstimatedK,distanceMoved,angleMoved,PreviousCovarianceOdometricPositionEstimate); + pc.printf("\n\r x:%f,y:%f,theta:%f",this->xWorld,this->yWorld,this->thetaWorld); + pc.printf("\n\r xEstim:%f,yEstim:%f,thetaEstim:%f",xEstimatedK,yEstimatedK,thetaEstimatedK); + } +} + + +//compute predicted localisation (assume movements are either straight lines or pure rotation), update PreviousCovarianceOdometricPositionEstimate +//TODO tweak constant rewritte it good +void MiniExplorerCoimbra::procedure_lab_4(float xEstimatedK,float yEstimatedK, float thetaWorldEstimatedK, float distanceMoved, float angleMoved, float PreviousCovarianceOdometricPositionEstimate[3][3]){ + + float distanceMovedByLeftWheel=distanceMoved/2; + float distanceMovedByRightWheel=distanceMoved/2; + if(angleMoved !=0){ + //TODO check that not sure + distanceMovedByLeftWheel=-angleMoved/(2*this->distanceWheels); + distanceMovedByRightWheel=angleMoved/(2*this->distanceWheels); + }else{ + distanceMovedByLeftWheel=distanceMoved/2; + distanceMovedByRightWheel=distanceMoved/2; + } + + float xEstimatedKNext=xEstimatedK+distanceMoved*cos(angleMoved); + float yEstimatedKNext=yEstimatedK+distanceMoved*sin(angleMoved); + float thetaEstimatedKNext=thetaWorldEstimatedK+angleMoved; + + float KRight=0.1;//error constant + float KLEft=0.1;//error constant + float motionIncrementCovarianceMatrix[2][2]; + motionIncrementCovarianceMatrix[0][0]=kRight*abs(distanceMovedRight); + motionIncrementCovarianceMatrix[0][1]=0; + motionIncrementCovarianceMatrix[1][0]=0; + motionIncrementCovarianceMatrix[1][1]=kLeft*abs(distanceMovedLeft); + + float jacobianFp[3][3]; + jacobianFp[0][0]=1; + jacobianFp[0][1]=0; + jacobianFp[0][2]=-distanceMoved*sin(thetaWorldEstimatedK+angleMoved/2); + jacobianFp[1][0]=0; + jacobianFp[1][1]=1; + jacobianFp[1][2]=distanceMoved*cos(thetaWorldEstimatedK+angleMoved/2);; + jacobianFp[2][0]=0; + jacobianFp[2][1]=0; + jacobianFp[2][2]=1; + + float jacobianFpTranspose[3][3]; + jacobianFpTranspose[0][0]=1; + jacobianFpTranspose[0][1]=0; + jacobianFpTranspose[0][2]=0; + jacobianFpTranspose[1][0]=0; + jacobianFpTranspose[1][1]=1; + jacobianFpTranspose[1][2]=0; + jacobianFpTranspose[2][0]=-distanceMoved*sin(thetaWorldEstimatedK+angleMoved/2); + jacobianFpTranspose[2][1]=distanceMoved*cos(thetaWorldEstimatedK+angleMoved/2); + jacobianFpTranspose[2][2]=1; + + float jacobianFDeltarl[3][2]; + jacobianFDeltarl[0][0]=(1/2)*cos(thetaWorldEstimatedK+angleMoved/2)-(distancedMoved/(2*this->distanceWheels))*sin(thetaWorldEstimatedK+angleMoved/2); + jacobianFDeltarl[0][1]=(1/2)*cos(thetaWorldEstimatedK+angleMoved/2)+(distancedMoved/(2*this->distanceWheels))*sin(thetaWorldEstimatedK+angleMoved/2); + jacobianFDeltarl[1][0]=(1/2)*sin(thetaWorldEstimatedK+angleMoved/2)+(distancedMoved/(2*this->distanceWheels))*cos(thetaWorldEstimatedK+angleMoved/2); + jacobianFDeltarl[1][1]=(1/2)*sin(thetaWorldEstimatedK+angleMoved/2)-(distancedMoved/(2*this->distanceWheels))*cos(thetaWorldEstimatedK+angleMoved/2); + jacobianFDeltarl[2][0]=1/this->distanceWheels; + jacobianFDeltarl[2][1]=-1/this->distanceWheels; + + float jacobianFDeltarlTranspose[2][3];//1,3,5;2,4,6 + jacobianFDeltarlTranspose[0][0]=(1/2)*cos(thetaWorldEstimatedK+angleMoved/2)-(distancedMoved/(2*this->distanceWheels))*sin(thetaWorldEstimatedK+angleMoved/2); + jacobianFDeltarlTranspose[0][1]=(1/2)*sin(thetaWorldEstimatedK+angleMoved/2)+(distancedMoved/(2*this->distanceWheels))*cos(thetaWorldEstimatedK+angleMoved/2); + jacobianFDeltarlTranspose[0][2]=1/this->distanceWheels; + jacobianFDeltarlTranspose[1][0]=(1/2)*cos(thetaWorldEstimatedK+angleMoved/2)+(distancedMoved/(2*this->distanceWheels))*sin(thetaWorldEstimatedK+angleMoved/2); + jacobianFDeltarlTranspose[1][1]=(1/2)*sin(thetaWorldEstimatedK+angleMoved/2)-(distancedMoved/(2*this->distanceWheels))*cos(thetaWorldEstimatedK+angleMoved/2); + jacobianFDeltarlTranspose[1][2]=-1/this->distanceWheels; + + float tempMatrix3_3[3][3]; + for(i = 0; i < 3; ++i){//mult 3*3, 3*3 + for(j = 0; j < 3; ++j){ + for(k = 0; k < 3; ++k){ + tempMatrix3_3[i][j] += jacobianFp[i][k] * PreviousCovarianceOdometricPositionEstimate[k][j]; + } + } + } + float sndTempMatrix3_3[3][3]; + for(i = 0; i < 3; ++i){//mult 3*3, 3*3 + for(j = 0; j < 3; ++j){ + for(k = 0; k < 3; ++k){ + sndTempMatrix3_3[i][j] += tempMatrix3_3[i][k] * jacobianFpTranspose[k][j]; + } + } + } + float tempMatrix3_2[3][2]; + for(i = 0; i < 3; ++i){//mult 3*2 , 2,2 + for(j = 0; j < 2; ++j){ + for(k = 0; k < 2; ++k){ + tempMatrix3_2[i][j] += jacobianFDeltarl[i][k] * motionIncrementCovarianceMatrix[k][j]; + } + } + } + float thrdTempMatrix3_3[3][3]; + for(i = 0; i < 3; ++i){//mult 3*2 , 2,3 + for(j = 0; j < 3; ++j){ + for(k = 0; k < 2; ++k){ + thrdTempMatrix3_3[i][j] += tempMatrix3_2[i][k] * jacobianFDeltarlTranspose[k][j]; + } + } + } + float newCovarianceOdometricPositionEstimate[3][3]; + for(i = 0; i < 3; ++i){//add 3*3 , 3*3 + for(j = 0; j < 3; ++j){ + newCovarianceOdometricPositionEstimate[i][j]=sndTempMatrix3_3[i][j]+thrdTempMatrix3_3[i][j]; + } + } + + //check measurements from sonars, see if they passed the validation gate + float leftCm = get_distance_left_sensor()/10; + float frontCm = get_distance_front_sensor()/10; + float rightCm = get_distance_right_sensor()/10; + //if superior to sonar range, put the value to sonar max range + 1 + if(leftCm > this->sonarLeft.maxRange) + leftCm=this->sonarLeft.maxRange; + if(frontCm > this->sonarFront.maxRange) + frontCm=this->sonarFront.maxRange; + if(rightCm > this->sonarRight.maxRange) + rightCm=this->sonarRight.maxRange; + //get the estimated measure we should have according to our knowledge of the map and the previously computed localisation + float leftCmEstimated=this->sonarLeft.maxRange; + float frontCmEstimated=this->sonarFront.maxRange; + float rightCmEstimated=this->sonarRight.maxRange; + float xWorldCell; + float yWorldCell; + float currDistance; + float xClosetCellLeft; + float yClosetCellLeft; + float xClosetCellFront; + float yClosetCellFront; + float xClosetCellRight; + float yClosetCellRight; + for(int i=0;i<this->map.nbCellWidth;i++){ + for(int j=0;j<this->map.nbCellHeight;j++){ + //check if occupied, if not discard + if(this->map.get_proba_cell(i,j)==1){ + //check if in sonar range + xWorldCell=this->map.cell_width_coordinate_to_world(i); + yWorldCell=this->map.cell_height_coordinate_to_world(j); + //check left + currDistance=this->sonarLeft.isInRange(xWorldCell,yWorldCell,xEstimatedKNext,yEstimatedKNext,thetaEstimatedKNext); + if((currDistance < this->sonarLeft.maxRange) && currDistance!=-1) + //check if distance is lower than previous, update lowest if so + if(currDistance < leftCmEstimated){ + leftCmEstimated= currDistance; + xClosetCellLeft=xWorldCell; + yClosetCellLeft=yWorldCell; + } + } + //check front + currDistance=this->sonarFront.isInRange(xWorldCell,yWorldCell,xEstimatedKNext,yEstimatedKNext,thetaEstimatedKNext); + if((currDistance < this->sonarFront.maxRange) && currDistance!=-1) + //check if distance is lower than previous, update lowest if so + if(currDistance < frontCmEstimated){ + frontCmEstimated= currDistance; + xClosetCellFront=xWorldCell; + yClosetCellFront=yWorldCell; + } + } + //check right + currDistance=this->sonarRight.isInRange(xWorldCell,yWorldCell,xEstimatedKNext,yEstimatedKNext,thetaEstimatedKNext); + if((currDistance < this->sonarRight.maxRange) && currDistance!=-1) + //check if distance is lower than previous, update lowest if so + if(currDistance < rightCmEstimated){ + rightCmEstimated= currDistance; + xClosetCellRight=xWorldCell; + yClosetCellRight=yWorldCell; + } + } + } + } + } + //get the innoncation: the value of the difference between the actual measure and what we anticipated + float innovationLeft=leftCm-leftCmEstimated; + float innovationFront=frontCm-frontCmEstimated; + float innovationRight=-rightCm-rightCmEstimated; + //compute jacobian of observation + float jacobianOfObservationLeft[3]; + float jacobianOfObservationRight[3]; + float jacobianOfObservationFront[3]; + float xSonarLeft=xEstimatedKNext+this->sonarLeft.distanceX; + float ySonarLeft=yEstimatedKNext+this->sonarLeft.distanceY; + //left + jacobianOfObservationLeft[0]=(xSonarLeft-xClosestCellLeft)/leftCmEstimated; + jacobianOfObservationLeft[1]=(ySonarLeft-yClosestCellLeft)/leftCmEstimated; + jacobianOfObservationLeft[2]=(xClosestCellLeft-xSonarLeft)*(xSonarLeft*sin(thetaEstimatedKNext)+ySonarLeft*cos(thetaEstimatedKNext)+(yClosestCellLeft-ySonarLeft)*(-xSonarLeft*cos(thetaEstimatedKNext)+ySonarLeft*sin(thetaEstimatedKNext)); + //front + float xSonarFront=xEstimatedKNext+this->sonarFront.distanceX; + float ySonarFront=yEstimatedKNext+this->sonarFront.distanceY; + jacobianOfObservationFront[0]=(xSonarFront-xClosestCellFront)/FrontCmEstimated; + jacobianOfObservationFront[1]=(ySonarFront-yClosestCellFront)/FrontCmEstimated; + jacobianOfObservationFront[2]=(xClosestCellFront-xSonarFront)*(xSonarFront*sin(thetaEstimatedKNext)+ySonarFront*cos(thetaEstimatedKNext)+(yClosestCellFront-ySonarFront)*(-xSonarFront*cos(thetaEstimatedKNext)+ySonarFront*sin(thetaEstimatedKNext)); + //right + float xSonarRight=xEstimatedKNext+this->sonarRight.distanceX; + float ySonarRight=yEstimatedKNext+this->sonarRight.distanceY; + jacobianOfObservationRight[0]=(xSonarRight-xClosestCellRight)/RightCmEstimated; + jacobianOfObservationRight[1]=(ySonarRight-yClosestCellRight)/RightCmEstimated; + jacobianOfObservationRight[2]=(xClosestCellRight-xSonarRight)*(xSonarRight*sin(thetaEstimatedKNext)+ySonarRight*cos(thetaEstimatedKNext)+(yClosestCellRight-ySonarRight)*(-xSonarRight*cos(thetaEstimatedKNext)+ySonarRight*sin(thetaEstimatedKNext)); + + float innovationCovarianceLeft[3][3]; + float innovationCovarianceFront[3][3]; + float innovationCovarianceRight[3][3]; + //left + float TempMatrix3_3InnovationLeft[3]; + TempMatrix3_3InnovationLeft[0]=jacobianOfObservationLeft[0]*newCovarianceOdometricPositionEstimate[0][0]+jacobianOfObservationLeft[1]*newCovarianceOdometricPositionEstimate[1][0]+jacobianOfObservationLeft[2]*newCovarianceOdometricPositionEstimate[2][0]; + TempMatrix3_3InnovationLeft[1]=jacobianOfObservationLeft[0]*newCovarianceOdometricPositionEstimate[0][1]+jacobianOfObservationLeft[1]*newCovarianceOdometricPositionEstimate[1][1]+jacobianOfObservationLeft[2]*newCovarianceOdometricPositionEstimate[2][1]; + TempMatrix3_3InnovationLeft[2]jacobianOfObservationLeft[0]*newCovarianceOdometricPositionEstimate[0][2]+jacobianOfObservationLeft[1]*newCovarianceOdometricPositionEstimate[1][2]+jacobianOfObservationLeft[2]*newCovarianceOdometricPositionEstimate[2][2]; + float innovationConvarianceLeft=TempMatrix3_3InnovationLeft[0]*jacobianOfObservationLeft[0]+TempMatrix3_3InnovationLeft[1]*jacobianOfObservationLeft[1]+TempMatrix3_3InnovationLeft[2]*jacobianOfObservationLeft[2]; + //front + float TempMatrix3_3InnovationFront[3]; + TempMatrix3_3InnovationFront[0]=jacobianOfObservationFront[0]*newCovarianceOdometricPositionEstimate[0][0]+jacobianOfObservationFront[1]*newCovarianceOdometricPositionEstimate[1][0]+jacobianOfObservationFront[2]*newCovarianceOdometricPositionEstimate[2][0]; + TempMatrix3_3InnovationFront[1]=jacobianOfObservationFront[0]*newCovarianceOdometricPositionEstimate[0][1]+jacobianOfObservationFront[1]*newCovarianceOdometricPositionEstimate[1][1]+jacobianOfObservationFront[2]*newCovarianceOdometricPositionEstimate[2][1]; + TempMatrix3_3InnovationFront[2]jacobianOfObservationFront[0]*newCovarianceOdometricPositionEstimate[0][2]+jacobianOfObservationFront[1]*newCovarianceOdometricPositionEstimate[1][2]+jacobianOfObservationFront[2]*newCovarianceOdometricPositionEstimate[2][2]; + float innovationConvarianceFront=TempMatrix3_3InnovationFront[0]*jacobianOfObservationFront[0]+TempMatrix3_3InnovationFront[1]*jacobianOfObservationFront[1]+TempMatrix3_3InnovationFront[2]*jacobianOfObservationFront[2]; + //right + float TempMatrix3_3InnovationRight[3]; + TempMatrix3_3InnovationRight[0]=jacobianOfObservationRight[0]*newCovarianceOdometricPositionEstimate[0][0]+jacobianOfObservationRight[1]*newCovarianceOdometricPositionEstimate[1][0]+jacobianOfObservationRight[2]*newCovarianceOdometricPositionEstimate[2][0]; + TempMatrix3_3InnovationRight[1]=jacobianOfObservationRight[0]*newCovarianceOdometricPositionEstimate[0][1]+jacobianOfObservationRight[1]*newCovarianceOdometricPositionEstimate[1][1]+jacobianOfObservationRight[2]*newCovarianceOdometricPositionEstimate[2][1]; + TempMatrix3_3InnovationRight[2]jacobianOfObservationRight[0]*newCovarianceOdometricPositionEstimate[0][2]+jacobianOfObservationRight[1]*newCovarianceOdometricPositionEstimate[1][2]+jacobianOfObservationRight[2]*newCovarianceOdometricPositionEstimate[2][2]; + float innovationConvarianceRight=TempMatrix3_3InnovationRight[0]*jacobianOfObservationRight[0]+TempMatrix3_3InnovationRight[1]*jacobianOfObservationRight[1]+TempMatrix3_3InnovationRight[2]*jacobianOfObservationRight[2]; + + //check if it pass the validation gate + float gateScoreLeft=innovationLeft*innovationLeft/innovationConvarianceLeft; + float gateScoreFront=innovationFront*innovationFront/innovationConvarianceFront; + float gateScoreRight=innovationRight*innovationRight/innovationConvarianceRight; + int leftPassed=0; + int frontPassed=0; + int rightPassed=0; + int e=10;//constant + if(gateScoreLeft < e) + leftPassed=1; + if(gateScoreFront < e) + frontPassed=10; + if(gateScoreRight < e) + rightPassed=100; + //for those who passed + //compute composite innovation + float compositeInnovation[3]; + int nbPassed=leftPassed+frontPassed+rightPassed; + switch(nbPassed){ + case 0://none + compositeInnovation[0]=0; + compositeInnovation[1]=0; + compositeInnovation[2]=0; + break; + case 1://left + compositeInnovation[0]=innovationLeft; + compositeInnovation[1]=0; + compositeInnovation[2]=0; + break; + case 10://front + compositeInnovation[0]=0; + compositeInnovation[1]=innovationFront; + compositeInnovation[2]=0; + break; + case 11://left and front + compositeInnovation[0]=innovationLeft; + compositeInnovation[1]=innovationFront; + compositeInnovation[2]=0; + break; + case 100://right + compositeInnovation[0]=0; + compositeInnovation[1]=0; + compositeInnovation[2]=innovationRight; + break; + case 101://right and left + compositeInnovation[0]=innovationLeft; + compositeInnovation[1]=0; + compositeInnovation[2]=innovationRight; + break; + case 110://right and front + compositeInnovation[0]=0; + compositeInnovation[1]=innovationFront; + compositeInnovation[2]=innovationRight; + break + case 111://right front and left + compositeInnovation[0]=innovationLeft; + compositeInnovation[1]=innovationFront; + compositeInnovation[2]=innovationRight; + break; + } + //compute composite measurement Jacobian + switch(nbPassed){ + case 0://none + break; + case 1://left + //compositeMeasurementJacobian = jacobianOfObservationLeft + break; + case 10://front + //compositeMeasurementJacobian = jacobianOfObservationFront + break; + case 11://left and front + float compositeMeasurementJacobian[6] + compositeMeasurementJacobian[0]= jacobianOfObservationLeft[0]; + compositeMeasurementJacobian[1]= jacobianOfObservationLeft[1]; + compositeMeasurementJacobian[2]= jacobianOfObservationLeft[2]; + compositeMeasurementJacobian[3]= jacobianOfObservationFront[0]; + compositeMeasurementJacobian[4]= jacobianOfObservationFront[1]; + compositeMeasurementJacobian[5]= jacobianOfObservationFront[2]; + break; + case 100://right + //compositeMeasurementJacobian = jacobianOfObservationRight + break; + case 101://right and left + float compositeMeasurementJacobian[6] + compositeMeasurementJacobian[0]= jacobianOfObservationLeft[0]; + compositeMeasurementJacobian[1]= jacobianOfObservationLeft[1]; + compositeMeasurementJacobian[2]= jacobianOfObservationLeft[2]; + compositeMeasurementJacobian[3]= jacobianOfObservationRight[0]; + compositeMeasurementJacobian[4]= jacobianOfObservationRight[1]; + compositeMeasurementJacobian[5]= jacobianOfObservationRight[2]; + + break; + case 110://right and front + float compositeMeasurementJacobian[6] + compositeMeasurementJacobian[0]= jacobianOfObservationFront[0]; + compositeMeasurementJacobian[1]= jacobianOfObservationFront[1]; + compositeMeasurementJacobian[2]= jacobianOfObservationFront[2]; + compositeMeasurementJacobian[3]= jacobianOfObservationRight[0]; + compositeMeasurementJacobian[4]= jacobianOfObservationRight[1]; + compositeMeasurementJacobian[5]= jacobianOfObservationRight[2]; + + break + case 111://right front and left + float compositeMeasurementJacobian[9] + compositeMeasurementJacobian[0]= jacobianOfObservationLeft[0]; + compositeMeasurementJacobian[1]= jacobianOfObservationLeft[1]; + compositeMeasurementJacobian[2]= jacobianOfObservationLeft[2]; + compositeMeasurementJacobian[3]= jacobianOfObservationFront[0]; + compositeMeasurementJacobian[4]= jacobianOfObservationFront[1]; + compositeMeasurementJacobian[5]= jacobianOfObservationFront[2]; + compositeMeasurementJacobian[6]= jacobianOfObservationRight[0]; + compositeMeasurementJacobian[7]= jacobianOfObservationRight[1]; + compositeMeasurementJacobian[8]= jacobianOfObservationRight[2]; + break; + } + //compute composite innovation covariance TODO dont have time, I assume you can just multiply them, can also simply all this easily + float compositeInnovationCovariance + switch(nbPassed){ + case 0://none + compositeInnovationCovariance=1; + break; + case 1://left + compositeInnovationCovariance = innovationConvarianceLeft; + + break; + case 10://front + compositeInnovationCovariance = innovationConvarianceFront; + break; + case 11://left and front + compositeInnovationCovariance=innovationConvarianceLeft*innovationConvarianceFront; + + break; + case 100://right + compositeInnovationCovariance = innovationConvarianceRight; + break; + case 101://right and left + compositeInnovationCovariance=innovationConvarianceLeft*innovationConvarianceRight; + + break; + case 110://right and front + compositeInnovationCovariance=innovationConvarianceFront*innovationConvarianceRight; + + + break + case 111://right front and left + compositeInnovationCovariance=innovationConvarianceLeft*innovationConvarianceFront*innovationConvarianceRight; + + break; + } + + //compute kalman gain + switch(nbPassed){ + //mult nbSonarPassed*3 , 3*3 + case 0://none + break; + case 1://left + float kalmanGain[3][3]; + for(i = 0; i < 3; ++i){//mult 3*3, 3*1 + for(j = 0; j < 1; ++j){ + for(k = 0; k < 3; ++k){ + kalmanGain[i][j] += newCovarianceOdometricPositionEstimate[i][k] * compositeInnovationCovariance[k]; + } + } + } + + break; + case 10://front + float kalmanGain[3][3]; + for(i = 0; i < 3; ++i){//mult 3*3, 3*1 + for(j = 0; j < 1; ++j){ + for(k = 0; k < 3; ++k){ + kalmanGain[i][j] += newCovarianceOdometricPositionEstimate[i][k] * compositeInnovationCovariance[k]; + } + } + } + break; + case 11://left and front + float kalmanGain[3][3]; + for(i = 0; i < 3; ++i){//mult 3*3, 3*2 + for(j = 0; j < 2; ++j){ + for(k = 0; k < 3; ++k){ + kalmanGain[i][j] += newCovarianceOdometricPositionEstimate[i][k] * compositeInnovationCovariance[k+j*3]; + } + } + } + + break; + case 100://right + float kalmanGain[3][3]; + for(i = 0; i < 3; ++i){//mult 3*3, 3*1 + for(j = 0; j < 1; ++j){ + for(k = 0; k < 3; ++k){ + kalmanGain[i][j] += newCovarianceOdometricPositionEstimate[i][k] * compositeInnovationCovariance[k]; + } + } + } + break; + case 101://right and left + float kalmanGain[3][3]; + for(i = 0; i < 3; ++i){//mult 3*3, 3*2 + for(j = 0; j < 2; ++j){ + for(k = 0; k < 3; ++k){ + kalmanGain[i][j] += newCovarianceOdometricPositionEstimate[i][k] * compositeInnovationCovariance[k+j*3]; + } + } + } + + break; + case 110://right and front + float kalmanGain[3][3]; + for(i = 0; i < 3; ++i){//mult 3*3, 3*2 + for(j = 0; j < 2; ++j){ + for(k = 0; k < 3; ++k){ + kalmanGain[i][j] += newCovarianceOdometricPositionEstimate[i][k] * compositeInnovationCovariance[k+j*3]; + } + } + } + + + break + case 111://right front and left + float kalmanGain[3][3]; + for(i = 0; i < 3; ++i){//mult 3*3, 3*3 + for(j = 0; j < 3; ++j){ + for(k = 0; k < 3; ++k){ + kalmanGain[i][j] += newCovarianceOdometricPositionEstimate[i][k] * compositeInnovationCovariance[k+j*3]; + } + } + } + + break; + } + for(i = 0; i < 3; ++i){//mult 3*3, 3*3 + for(j = 0; j < 3; ++j){ + kalmanGain[i][j] = kalmanGain[i][j][i][k]/compositeInnovationCovariance; + } + } + //compute kalman gain * composite innovation + //mult 3*3 , 3*1 + float result3_1[3][1]; + for(i = 0; i < 3; ++i){//mult 3*3, 3*1 + for(j = 0; j < 1; ++j){ + for(k = 0; k < 3; ++k){ + result3_1[i][j] += kalmanGain[i][k] * compositeInnovation[k]; + } + } + } + //compute updated robot position estimate + float xEstimatedKLast=xEstimatedKNext+result3_1[0]; + float yEstimatedKLast=yEstimatedKNext+result3_1[1]; + float thetaEstimatedKLast=thetaEstimatedKNext+result3_1[2]; + //store the resultant covariance for next estimation + /* + a b c + d e f + g h i + + a d g + b e h + c f i + */ + float kalmanGainTranspose[3][3]; + kalmanGainTranspose[0][0]=kalmanGain[0][0]; + kalmanGainTranspose[0][1]=kalmanGain[1][0]; + kalmanGainTranspose[0][2]=kalmanGain[2][0]; + kalmanGainTranspose[1][0]=kalmanGain[0][1]; + kalmanGainTranspose[1][1]=kalmanGain[1][1]; + kalmanGainTranspose[1][2]=kalmanGain[2][1]; + kalmanGainTranspose[2][0]=kalmanGain[0][2]; + kalmanGainTranspose[2][1]=kalmanGain[1][2]; + kalmanGainTranspose[2][2]=kalmanGain[2][2]; + + //mult 3*3 , 3*3 + float fithTempMatrix3_3[3][3]; + for(i = 0; i < 3; ++i){//mult 3*3 , 3*3 + for(j = 0; j < 3; ++j){ + for(k = 0; k < 3; ++k){ + fithTempMatrix3_3[i][j] += kalmanGain[i][k] * kalmanGainTranspose[k][j]; + } + } + } + for(i = 0; i < 3; ++i){//add 3*3 , 3*3 + for(j = 0; j < 3; ++j){ + fithTempMatrix3_3[i][j]=fithTempMatrix3_3[i][j]*compositeInnovationCovariance; + } + } + float covariancePositionEsimatedLast[3][3]; + for(i = 0; i < 3; ++i){//add 3*3 , 3*3 + for(j = 0; j < 3; ++j){ + covariancePositionEsimatedLast[i][j]=fithTempMatrix3_3[i][j]+newCovarianceOdometricPositionEstimate[i][j]; + } + } + //update PreviousCovarianceOdometricPositionEstimate + for(i = 0; i < 3; ++i){ + for(j = 0; j < 3; ++j){ + PreviousCovarianceOdometricPositionEstimate[i][j]=covariancePositionEsimatedLast[i][j]; + } + } + + xEstimatedK=xEstimatedKLast; + yEstimatedK=yEstimatedKLast; + thetaEstimatedK=thetaEstimatedKLast; +}