autumnday / Model-Predictive-Control

This package contains implementation for plan synthesis algorithms given a finite transition system (as the agent motion model) and a Linear temporal logic formula (as the agent task). It outputs the online path plan as a sequence of agent motion and action, required to fulfill the task.

Description

1. This package contains implementations for online-plan synthesis algorithm give a dynamic environment (as the finite transition system) and a potential infeasible Linear temporal logic formula (as the robot’s task). It outputs the predicted trajectories at each time-step that optimize the reward-related objectives and finally fulfill the task maximumly.

• Simulation Video: Simulation video

2. In the experiment, the neighbor numbers of simulation snapshot provide the time-varying reward (Here we use the random values) which is our optimization objective.

• Experiment Video: Experiment video

Reference

Receding Horizon Control Based Online LTL Motio sPlanning in Partially Infeasible Environments. Mingyu Cai, H. Peng and Z. Kan. Journal of Autonomous Robot.paper link

Features

• Allow both normal and infeasible LTL based product automaton task formulas
• Motion model can be dynamic and initially unknown
• Soft specification is maximumly satisfied.
• Online-Path planning is designed from the model predicted control methodology.
• Collect and transfer the real-time data via Optitrack camera systems
• Allow automatically calibrate the mobile robots to obtain its orientation and dynamics.

Debugging

Ptthon3

• Install python packages like networkx2.0.ply
• Add to your PYTHONPATH, to import it in your own project.
• ltlba_32 and ltlba_64 are executable files complied under Linux, please follow [ltl2ba/README.txt]
• Try path_plan.py

Matlab

• Add folder to your PYTHONPATH.
• Try Receding_Horizon_Control.py