Interactive Motion Planning

Codebase for adaptive interactive mixed-integer model predictive control (aiMPC): an optimal control-based interactive motion planning algorithm for autonomous vehicles.

• Real-time Software-and-Human-in-the-loop simulation in CARLA.

Test scenario illustration

Mandatory lane change scenario: a stopped truck on the right lane necessitates a lane change for the autonomous vehicle which needs to negotiate with a human-driven vehicle on the left lane. - Blue vehicle is the ego vehicle and red vehicle is the human-driven neighboring vehicle (NV).

• aiMPC estimates NV's cost online and adapts the MPC.

A case when ego merges ahead

https://github.com/autonomous-viranjan/Interactive-Motion-Planning/assets/62226470/80b97b3e-c7b9-4cf5-933c-67992a033649

A case when ego merges behind

https://github.com/autonomous-viranjan/Interactive-Motion-Planning/assets/62226470/f22663f6-780c-471e-94e3-66f41bb8012c

A case where the NV's nature changes

https://github.com/autonomous-viranjan/Interactive-Motion-Planning/assets/62226470/bacb1f65-077f-4fc9-8c1b-87f18a4aafa8

• α_p and α_a are the estimated NV cost weights.

Key dependencies:

• CARLA simulator
• Gurobi
• ROS Noetic
• MATLAB (data analysis)

Architecture

• Cite as

@article{bhattacharyya2024automated,
  title={Automated Lane Change via Adaptive Interactive MPC: Human-in-the-Loop Experiments},
  author={Bhattacharyya, Viranjan and Vahidi, Ardalan},
  journal={IEEE Transactions on Control Systems Technology},
  year={2024},
  publisher={IEEE}
}