Perception based trajectory suggestion with machine learning
- Reference number
- ID18-0011
- Start and end dates
- 190101-240331
- Amount granted
- 2 500 000 SEK
- Administrative organization
- KTH - Royal Institute of Technology
- Research area
- Information, Communication and Systems Technology
Summary
Autonomous vehicles will have a huge impact impact on the society, including autonomous heavy trucks in less structured dynamic environments, like underground mines and complex construction sites. Holistic trajectory suggestion of possible drivable paths for heavy commercial vehicles in less structured and dynamic environments. - Finding and using key features from the environment using multiple sensors like camera, LiDAR, RADAR to understand the drivable path. - Developing a scientific approach for training deep neural networks. - Produce path candidate rankings in unstructured environments and help in identifying safe drivable path by exploring holistic path predictor and reinforcement learning. - Explore Transfer Learning techniques to understand other scenarios from known off-road scenarios. Expected results include methods and algorithms for perception based trajectory suggestion with machine learning. The expected results will be scientifically published.
Popular science description
Autonomous vehicles are going to have a huge impact on the society. However, to realize such a complex system there are many challenges in the area of perception, planning and control. The goal of the project is to deal with the challenges in perception domain for off-road scenarios with the help of machine learning techniques using different sensors like – Camera, LiDAR and Radar. The project will help to compute the possible path rankings and suggest suitable drivable path in off-road scenarios based on the raw data correlation between the aforementioned sensors. Another important objective of the project will be to find out the key features with which we can model different dynamic environments for autonomous driving. The techniques developed for driving in off-road scenarios will also be evaluated for driving autonomously in other dynamic environments since the Scania trucks operate in different environments.