Documents

Project Reports

  1. Conceptual Design Review (CoDR, December 2019)
  2. Critical Design Review (CDRR, May 2020)
  3. Final Report (December 2020)

Project Presentations

  1. Preliminary Design Presentation (PDR, March 2020)
  2. Spring Validation Demonstration Tests (SVD, April 2020)
  3. Critical Design Review Presentation (CDR, May 2020)
  4. System Development Review Presentation (SDR, October 2020)
  5. Standards and Regulations Presentation (S&R, December 2020)

Progress Reviews

  1. Progress review 1, presented by Rohan
  2. Progress review 2, presented by Sanil
  3. Progress review 3, presented by Poorva
  4. Progress review 4, presented by Uma
  5. Progress review 5, presented by Sachit
  6. Progress review 6, presented by Rohan
  7. Progress review 7, presented by Sanil
  8. Progress review 8, presented by Poorva
  9. Progress review 9, presented by Uma
  10. Progress review 10, presented by Sachit
  11. Progress review 11, dress rehearsal

Individual Lab Reports

  1. ILR-01: Poorva, Uma, Sachit, Sanil, Rohan
  2. ILR-02: Poorva, Uma, Sachit, Sanil, Rohan
  3. ILR-03: Poorva, Uma, Sachit, Sanil, Rohan
  4. ILR-04: Poorva, Uma, Sachit, Sanil, Rohan
  5. ILR-05: Poorva, Uma, Sachit, Sanil, Rohan
  6. ILR-06: Poorva, Uma, Sachit, Sanil, Rohan
  7. ILR-07: Poorva, Uma, Sachit, Sanil, Rohan
  8. ILR-08: Poorva, Uma, Sachit, Sanil, Rohan
  9. ILR-09: Poorva, Uma, Sachit, Sanil, Rohan
  10. ILR-10: Poorva, Uma, Sachit, Sanil, Rohan

Design Brainstorming

Check here

Software

Our code-base can be found on the team’s GitHub page.

Hardware

We are using PIXBOT, the world’s first open-source autonomous chassis, designed and developed by PIX Moving. The vehicle control code, CAD models and other design documentation are detailed on the Gitlab page here.

Sensor Datasheets

  1. Mindvision Industrial Camera MV-SUF401 GC
  2. Velodyne VLP-16 LiDAR Puck

PCB Design

To mount a LIDAR and the NVIDIA Jetson on the RC car, our temporary testbed, we would need to build a power distribution system to efficiently power all required components.

Power Source
The source of Electrical power for our system is a series combination of two 11.1V LiPo battery. This is because our components require a maximum of 12V and connecting in series will ensure we supply the required voltages at all times. The specifications of the battery are as follows:

  • Turnigy 1450mAh 3S 11.1v Transmitter Lipoly Pack
  • Size: 93x40x14mm
  • Charge Plug: JST-XH
  • Discharge: Futaba JR
  • The LiPo battery has a voltage range of 10.7 to 12.4V between charges and has a current output of 1.45 Amps

Powered Subsystems

  • Nvidia Jetson TX1
    It can take a maximum of 19V, 4A supply input. We are planning to supply a 12V/3A supply to the Jetson. Since our battery supply is equivalent to 22.2V, we would use the MIC29300-12WU voltage regulator. The peak current allowed for proper functioning of the Jetson is 4A, and we plan to operate within 3A. We will use 1 CGRID connector for supplying output to the Jetson. We will have a circuit for reverse voltage and over-current protection by including a diode and a fuse.
  • Lidar
    We will also have a 12V/2A output for the VLP16 Velodyne LIDAR. Since the current requirements are different compared to the Jetson, we will use a separate Voltage regulator for the same. Again, a CGRID connector will be used at the output and protection will be given against reverse input, short circuit and over voltage. An LED indicator would be used to indicate whether or not this sensor is powered or not. We will also add a manual switch for the Lidar at the output so that power is not distributed unnecessarily.

Figure 1: Power Distribution Board Schematic

Figure 2: PCB Power Distribution System Layout

Datasheets:

Voltage regulator (high current family): MIC 29300

Voltage reference diode: BZV10

Software Flowcharts

                                                                              
Pod Localization                                 Waypoint Generation                                  Approach Navigation                                                                          Position Verification

Component Evaluation and Testing

Most of the hardware component evaluation and tests will be taken up in the fall semester. We did the following experiments and testings in simulation.

Experiment Procedure Results
1. Evaluate use of 3D LiDAR for docking verification position Place the chassis under the pod at different orientation and test offset against ground truth We found that the 3D point cloud at certain channels of VLP-16 can be used to detect pod legs and we can find the offset from the center. However, the accuracy was not consistent due to VLP-16 simulation having some bugs observing the same data over time. We also found that VLP-16 data under 1m has lower accuracy and is more prone to noise. We will evaluate this on hardware 
2. Evaluate 2D LiDAR for Pod leg detection Place chassis in front of Pod in different orientations and estimate pod center We were able to estimate the center of Pod with high accuracy even when we were viewing 2 or 3 legs, it was very robust. However, if the range exceeded the LiDAR beams were missing pod legs. We used this to define our PHZ length.
3. Evaluate fiducial markers for Pod estimate  Place chassis in front of Pod in different orientations and place different markers on Pod, estimate position of markers. Change marker type and sizes We tested ARTags and AprilTags and decided to use the latter due to easy interfacing. The size of AprilTag was decided to be 20cm to accurately detect and estimate pose at PHZ length
4. Tune PD parameters for Pure Pursuit Navigation Give waypoints to chassis and evaluate accuracy at final waypoint We found that a damping term helped the chassis not to overshoot the target and tune parameters. We also found that we can add a potential field based constraints on velocity and steering angle based upon the Euclidean distance to the final waypoint to improve accuracy. It also made the system more reliable as in the chances of a collision near the final waypoint reduced.
5. Find optimum Placement of sensors Place sensors at various location and test their functionality and ease of accessibility

  1. 2D LiDAR – Pod Leg detection and Obstacle Detection
  2. 3D LiDAR – Localization and position verification
  3. Camera – Object detection and
 We observed that

  1. For 2D LiDAR, placement at front of the vehicle gives the best field of view as well as reliability in obstacle detection as we can detect obstacles farther away from chassis earlier. It detects Pod Legs till 0.3m away from docking point after which we can move using our estimate
  2. For 3D LiDAR – Ideal location is in the center of the vehicle as it gives a good 360-degree view, however, that will cause problems once our chassis has lifted the Pod
  3. For camera – the best Feild of View was found at the front of the vehicle