madhavkarri / carnd-p7-pathplanning

Self-Driving Car. Project: Highway Driving Path Planning

Self-Driving Car

Project: Highway Driving Path Planning

MK

Overview

Develop a path planner in c++ to safely navigate around a virtual 3 lane highway with other traffic that is driving +/-10 MPH of the 50 MPH speed limit. A successful path planner would keep inside its lane, avoid hitting other cars, and pass slower moving traffic using localization, sensor fusion, and map data.

The Project

The goals/steps for this project:

• Ego-car should maintain vehicle speed close to 50 mph, whenever possible
• Ego-car must avoid hitting other cars during all manuvers
• Ego-car must drive inside of the marked road lanes, except when changing lanes
• Ego-car must atleast complete one complete loop around the 6946 (m) length of highway
• Ego-car should take a little over 5 minutes to complete 1 loop
• Ego-car must maintain: (acceleration < 10 m/s^2) and (jerk: rate of change of acceleration < 10 m/s^3)

Files Submitted

1. Files included with submission to run the simulator in autonomous mode

Project includes the following files:

• main.cpp contains all the relevant, helper (helpers.h), and header files necessary to execute the path planner to control ego-car within the simulator
• helpers.h: functions "getXY" (transform from Frenet s,d coordinates to Cartesian x,y) and "distance" (Calculate distance between two points) are the two main functions defined within helpers.h
• spline.h: a function to construct a smooth spline (piece-wise polynomial) passing through a set of X and Y points. spline.h is used in developing a smooth trajectory generation
• README.md summarizes the results
• The first attempt resulted in a distance of 38.9 miles without incident. Final output video Link

2. Project code (main.cpp)

The main.cpp file contains code for the following set of tasks:

• Read/load map data to acquire highway waypoint data Link
• Connect to highway simulator using micro-websocket. Websocket handles two-way data transfer between planner and simulator

Sensing (Telemetry)

Gather below listed data from simulator Link:

• Ego car localization data
• Previous unused path from the simulator
• Localization data of other cars

Prediction Link:

• Based on data from sensing, predict other cars localization with respect to ego car. Detect whether other cars are present ahead, left, right of ego car

Behavior Planning and Trajectroy Generation Link:

Behavior Planning:

• If other car ahead of ego car (impement below listed actions in oder)
  • Check no other car on left lane and if left lane available, switch to left lane
  • If not, check no car on right lane and if right lane available, switch to right lane
  • Else, deccelerate ego car vehicle speed to avoid hitting other car ahead
• If ego car not on center lane, switch to center lane if it is safe to do so
• If ego car vehicle speed less than maximum allowable vehicle speed (47.5 mph), accelerate ego car vehicle speed

Trajectroy Generation

• In a nutshell, trajectory generation is performed by combining waypoints from previous path trajectory and a new set of way points generated by using a spline curve
• The inputs to generate a spline curve are as follows:
  • The last two waypoints from the previous path trajectory
  • To the above two waypoints 3 additonal waypoints are added at increments of 30, 60, and 90 meters
  • Addition of end points from previous path trajectory to the input of spline function will result in a smooth transition between successive path trajectories
• New path trajectory is generated as follows:
  • Step 1: Initialize new path trajectory by adding unused waypoints from previous path trajectory
  • Step 2: To step 1 add new waypoints by using spline curve generated previously
  • The combined total of previous and new waypoints in the new path trajectory is a fixed set of 50 waypoints
  • In addition, irrespective of the number of new waypoints added to the new path trajectory, the combined distance of the newly added waypoints is a fixed length of 30 meters

Next Steps and Future Updates

• In the presence of other car ahead of ego car, the ego-car does not smoothly follow the other car. Improvize the logic, may be maintain a constant pre-set distance to the other car
• May be explore to implement finite state machines