Collision detection and avoidance using extrapolation
Ajinkya Dobhada, Suraj Bhala, Pranjal Shah, Milind B Kamble · 2017
One of the major concerns in the existing collision avoidance systems is the accurate prediction of trajectory of a moving obstacle. Projection of the obstacle's motion is necessary for the collision detection well in advance. Hence in this paper we propose an efficient way of implementing the same by observing the motion of the moving obstacle for a certain interval of time. We then track the path of the obstacle using ultrasonic sensors, camera and servo motor. The tracked path is further extrapolated to determine the probability of collision of the obstacle and the user. The collision is detected by comparing the time required by the user and the obstacle to reach the possible point of intersection of the user and obstacle's path. We also define a confidence interval (in terms of time) to anticipate the possibility of collision well before in time so as to alert the user. This algorithm could be implemented in robotic guide system or in developing systems such as hand wearables or helmet that helps visually impaired or old age people in navigating through crowds.