P18542: Real Time Terrain Mapping
/public/

Integrated System Build & Test

Table of Contents

Team Vision for Integrated System Build & Test Phase

This phase our team focused on completing our individual components and working to integrate the components into a working system. These individual components were the vision system, search algorithm, robot and its firmware, as well as the software integration on the Pi.

Test Results Summary

Vision System

The vision system is nearly complete with only a few more tasks to complete. Right now, the system is able to take in an image from the raspberry pi, transform its perspective,and locate each terrain tile. After each tile is identified, the coordinates of the tiles are calculated along with the HSV color translated to RBG. After all of the attributes of the terrain tiles are identified, a json file is encoded to send over to the path finding algorithm to find the best path through it.
Feature Matching

Feature Matching

For tracking the robot, feature matching will be used. This matches key features of the robot to a stock image allowing us to positively identify it in the terrain. The orientation will be determined after the robot is identified by fitting a shape over the identified robot and finding the center line which we can then measure the angle from the axes of the terrain map.

Robot Construction

The robot was constructed for the main simulation. The initial tests showed that the wheels had a wobble and this was due to the bearings that we purchased. Since the bearings were not high quality, they were not good for support. However, the addition of a second bearing and mounting bracket greatly reduced the wobble. The axel shafts have been recut to satisfy this change. The modeling for the sensor holder needed to house the sensors for the snow simulation is going through its final revisions to maximize the effectiveness of the IR receiver.
Constructed Robot

Constructed Robot

Robot Firmware

Robot Firmware

Robot Firmware

Pi Software Integration

The Raspberry Pi has now been setup with all packages required to run our software and to host it's own Wifi access point. A number of applications have been successfully tested in isolation on the Pi's environment including the web server, vision algorithm, camera engine, communication server, and the database interface.

Work Remaining:

Search Algorithm

The path-finding algorithm was updated to be compatible with the test procedures. Three heuristics were designed for autonomous navigation. One of the heuristics determined the fastest path while ignoring the terrain. The second was generated to identify the safest path to simulate the protection of first responders within the ASAR vehicle. For example, the vehicle will not travel through fire while operating in safe mode. As a compromise, the third heuristic balances safety and urgency by following the safest and quickest general path. This heuristic allows the ASAR Unit to travel through the fire terrain to rescue a victim in critical condition. The resulting distinct paths are displayed below where the red path is the fastest path, the green is the safest, and the blue path is the compromise.
Best-Path Example

Best-Path Example

Best-Path Example

Best-Path Example

Testing was started on the algorithm. The JSON file containing the terrain information was successfully decoded after being received from the vision system. To complete this test satisfactorily the orientation of the robot will also need to be sent and successfully decoded.

The admissibility and consistency tests for the algorithm cannot be completed until the simulation terrain is constructed. Preliminary tests in this area have been completed and led to tuning and other small adjustments to improve the performance of the algorithm.

The function converting the identified best path to movement instructions for the robot was successful. The instructions are output as a single string to be sent wirelessly as a string of bits to the ASAR Unit. Additional testing is needed to ensure the instructions result in correct movements in the robot.

Simulation Terrain

The terrain has not been started. It has all been tested through images created in paint or printed images. This terrain construction will be started at the conclusion of this phase for further testing through the Raspberry Pi camera. It will be made of 2x4 wood and matte corrugated plastic to prevent reflection from light. A polarized filter was purchased to reduce this effect.

Risk and Problem Tracking

Risk Assessment

Test Plans

Plans for next phase

As the end of the school year and Senior Design are coming quickly to an end, the team hopes to have full integration completed as soon as possible to have a working prototype for Imagine RIT. This will also allow us to work out any kinks in our sytstem before the Imagine deadline.

Home | Planning & Execution | Imagine RIT

Problem Definition | Systems Design | Preliminary Detailed Design | Detailed Design

Build & Test Prep | Subsystem Build & Test | Integrated System Build & Test | Customer Handoff & Final Project Documentation