Human ground robot interaction
| Title | Human ground robot interaction |
|---|---|
| Summary | External communication from mobile robots to minimize conflicts with pedestrians |
| Keywords | Deep learning, Intention recognition, behavior prediction |
| TimeFrame | |
| References | |
| Prerequisites | Machine learning |
| Author | |
| Supervisor | Cristofer Englund, Martin Cooney, Fernando Alonso-Fernandez |
| Level | Master |
| Status | Open |
Self-driving mobile robots have become a hot alternative in last-mile delivery services. To make mobile robots socially acceptable, their behavior should be interpretable and predictable by humans (and other robots). This project will investigate how robot behavior can influence the behavior of pedestrians. The mobile robot should detect a pedestrian, and be programmed in such a way that the pedestrian should easily understand its intention i.e. if it will stop to let the pedestrian pass, or if it will be taking the initiative to make the pedestrian stop.
There are two robots available from Clearpath (https://clearpathrobotics.com/jackal-small-unmanned-ground-vehicle/), that can be used for the experiments. The idea is to use recurrent neural networks or LSTM to learn the different behaviors (https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7780479). Reinforcement learning is another method to take on the identification of behavior and movements. In (https://doi.org/10.3389/frai.2021.550030) a review of explainable AI and Reinforcement Learning is presented that can give inspiration to the given problem. Recent advances in transformers may also be a plausible approach (https://openaccess.thecvf.com/content/CVPR2021/html/Liu_Multimodal_Motion_Prediction_With_Stacked_Transformers_CVPR_2021_paper.html)
Does this sound interesting?
Contact us for more information.
