Cooperative Perception of Connected Vehicles for Safety

In cooperative perception, reliably detecting surrounding objects and communicating the information between vehicles is necessary for safety. However, vehicle-to-vehicle transmission of huge datasets or images can be computationally expensive and often not feasible in real time. A robust approach to ensure cooperation involves relative pose estimation between two vehicles sharing a common field of view. Detecting the object and transferring its location information in real time is necessary when the object is not in the ego vehicle’s field of view. In such scenarios, reliable and robust pose recovery of the object at each instant ensures the ego vehicle accurately estimates its trajectory. Once pose recovery is established, the object’s location information can be obtained for future trajectory prediction. Deterministic predictions provide only point estimates of future states which is not trustworthy under dynamic traffic scenarios. Estimating the uncertainty associated with the predicted states with a certain level of confidence can lead to robust path planning. This study proposed quantifying this uncertainty during forecasting using stochastic approximation, which deterministic approaches fail to capture. The current method is simple and applies Bayesian approximation during inference to standard neural network architectures for estimating uncertainty. The predictions between the probabilistic neural network models were compared with the standard deterministic models. The results indicate that the mean predicted path of probabilistic models was closer to the ground truth when compared with the deterministic prediction. The study has been extended to multiple datasets, providing a comprehensive comparison for each model.