path-planning

Collision avoidance path planning project for robots.

Prerequisite

1. CasADi

Quick Start

1. from navigator import MPC
2. Inputs of MPC class:
   • current state of ego robot
   • current state & predicted future states of surrounding agents
3. Output of MPC class:
   • current control action of ego robot

Useful Notes

1. Robot state vector contains x, y. Action vector contains vx, vy.

2. Properties of MPC class:

   • lookahead_step_num: The prediction horizon for MPC. Default is 5.
   • lookahead_step_timeinterval: The time interval between two consecutive states planned by MPC. Default is 0.1s. Make sure this equals to the time interval in your simulation.
   • end_point: The goal state of ego robot.
   • num_of_agent: The number of surrounding agents.
   • safety_r: The minimum distance allowed between ego robot and surrounding agents.
   • max_v: The maximum velocity allowed for ego robot. We assume max_vx=max_vy=max_v. Default is 0.3m/s.

3. Call function Solve(state, agent_state_pred) repeatedly in your main loop. Arguments:

   • state: The current state of ego robot. This should be a 1-D list [ego_x0, ego_y0].
   • agent_state_pred: The current state and predicted future states of surrounding agents. This should be a 3-D list [[[agent_1_x0, agent_1_y0], [agent_1_x1, agent_1_y1], ...], [[agent_2_x0, agent_2_y0], [agent_2_x1, agent_2_y1], ...], ...]. The shape of this list is determined by num_of_agent and lookahead_step_num.

4. Adjustable parameters in MPC class:

   • w_cte: The larger, the faster ego robot moves towards the goal state.
   • w_dv: The larger, the smoother ego robot trajectory is.