Autonomous Land Vehicle

This page provides a basic overview of the equipment used in my first robotics project: The Red Hawk 1 Autonomous Land Vehicle. I designed and built the RH1 during my senior undergraduate year at Montclair State University. Attached at the bottom of the page is the proposal I submitted to Mathematical Sciences and Computer Science departments to get funding for an autonomous vehicle project. There is also the paper I submitted afterwards discussing some of the software and hardware used by the vehicle.

Special Thanks

I just want to give special thanks to the Mathematical Sciences and Computer Science Departments at MSU for giving the "OK" on this project and funding it. Otherwise it would have never happened. I would also like to thank the North Jersey Astronomical Group for their donation and Tru Manufacturing for providing us with discounted machine work. Many thanks to EZ-Go as well, for donating the RXV golf cart straight from their factory in Georgia.

Processing and Controls

• Keyboard, mouse, and a CH Products industrial joystick all on swivel mounts.
  • Joystick used in Manual Control Mode (MCM) for steering, braking, and acceleration.
  • Joystick button used to switch in and out of MCM
• Dual 3.33Ghz Intel Xeon desktop running Ubuntu 10.04

3D Sensing

• Two Sick LMS 211 line scanners
  • The LIDARs provided an RS-485 signal that was opto-isolate and converted to USB
• Each LIDAR is oscillated by a high torque NEMA 34 stepper motor with US Digital optical quadrature encoders for position feedback
  • Oscillation position, laser angle, and laser 'time of flight' provide a front view point cloud
• All cabling is industrial grade, double shielded, high flex
• A micro switch is mounted on the motor pulley for homing
• Frame is 6061 aluminum (special thanks to TRU Manufacturing in Northvale, NJ for providing discounted machining)
  • Each LIDAR is sandwiched between two high load thrust bearings
• A custom shock mounted square tube frame supports both oscillating units

Motion Control

• Three Gecko stepper drivers for the Oscillators and Steering Control
• Mounted in isolated aluminum enclosure for adequate RFI shielding

• Motion control provided by a Galil Controller (left) over an ethernet connection
  • Some DB extensions made for a quick breakout setup
• Older system of the right is a JrKerr motion control network. Couldn't provide real time position updating needed by steering control. This was a "live and learn" scenario after meticulous custom cabling to provide proper breakouts.

Localisation

• Localisation was provided by a Kearfott VRU. This is a military grade Inertial Measurement System. Using a laser ring gyroscope to provide very accurate positioning with minimal drift.
• Like the LIDAR, this unit provided an RS-485 output which was opto-isolated and converted to USB.
• Provided accurate results for 3-5 minutes before drift set in.

Drive by Wire

• Controls the vehicles electronic braking and acceleration controls.
• The vehicles pedals use switched hall effect sensors to generate a voltage between 0.5 and 4.5VDC for the vehicles speed controller.
• Local power regulation of 5VDC and 9VDC.
• Two LTC1257 12-bit DACS wired in serial provide the 0.5 to 4.5VDC output to simulate the hall effect sensors used in the pedals.
• A relay board that simulates the switches mounted on the brake and accelerator pedals.
• A Parallax Basic Stamp that controls the relays, DACS, and communicates with the PC over an opto-isolated MAX232 chip.
  • Should the micro-controller not receive communication from the PC for a predetermined time, the controller enters an emergency stop sequence and brings the vehicle to an immediate stop.
• A microcontroller reset button, power switch, LED power indicator, and emergency stop LED indicator mounted on the enclosure
• UPDATE: This system is currently being updated to a CANopen controller that interfaces directly with the the RXV motor controller on the golf cart to give a more "elegant" solution in the control of the brake, accelerator, forward, and reverse controls.

Power

• The heart of the power distribution was a Honda EU2000i generator that provided true sine wave 110VAC (see pictures above)
• Various fused DC outputs were distributed to all electronics
  • Unregulated 80VDC for the stepper drivers
  • 24VDC for the Kearfott VRU
  • 5VDC and 12VDC for the 'drive by wire' electronics
• The entire vehicle and all electronics were given a common ground.
  • The ground is tied to a collection of small chains that dangle on the ground to sink any static charge to Earth.

First test : Obstacle avoidance

• The cart takes a slight right turn at start to get on the waypoint path and once it reaches the garbage can obstacle it does a real time path reroute around it.
• The garbage can was about two feet into the path of the cart.

'Making Of' Photos

(The welder is Andrew Huth, a math major at MSU and the guy on the lathe is Dr. Bob Dorner, retired physics professor from MSU)

Point Cloud Visualization

These images are from the initial programming of the oscillator system. The first image is a color representation of a polar view from a single Lidar oscillation over about 30 seconds to allow all pixels to fill in. The second image is a Cartesian transformation of the polar view. The golf cart uses a "birds-eye" Cartesian view of the Lidar data and looks at the rate of change in surface height (i.e. the gradient). Large variations in surface height create large gradient values. The gradient map is loaded onto the GPS map. If a large gradient value is seen in the same location several times, the mapping system marks this location as occupying an obstacle. More specifically, the more an obstacle appears in a particular location, the higher the probability that there is actually an obstacle there. This helps remove anomaly obstacles from the map. The path planning system is then given a copy of this map and does a geometric 'run through' of the desired path on the obstacles map. If there is an obstacle in the current path, the path is shifted to avoid the obstacle.