Machine Learning Based Techniques for Path Prediction

Path prediction aims at producing safer systems by allowing them to anticipate the outcomes of road scenes situations. Lately, machine learning methods have been used extensively for that purpose. Neural networks in particular with architectures such as RNN, LSTM, CNN, and self-attention. They offer the best results with the commonly used metrics. However, these evaluation criteria do not guarantee safety and can be criticized. More requirements should be met than the minimization of a few metrics.

Oct 22, 2019