Preliminary Design Review
Poorva, Sanil, Sachit, Rohan, Uma
Team J: Align
1
Project Description
2
PIXBOT Chassis
Frame representative of Pod
Use Case
3
4
5
6
7
8
9
10
11
12
13
14
15
!
16
17
Mandatory Performance Requirements
The system will:
M.P.1 travel with a maximum speed of 20 kmph with a payload.
M.P.2 plan a path of distance less than 120 km between charges.
M.P.3 travel with a maximum speed of 25 kmph without a payload.
M.P.4 identify the payload handling zone (PHZ) when the chassis is at a distance of 2 m from the periphery of PHZ.
M.P.5 complete docking in less than 120 s. This includes docking initiation, verification and retries.
M.P.6 align with the docking path with an error margin of ± 2 cm.
M.P.7 initiate communication and update sensor and vehicle models with a maximum latency of 60 s.
M.P.8 diagnose failures within 2000 ms from occurrence.
M.P.9 detect object location with an accuracy of ± 5cm.
M.P.10 maintain a minimum distance of 30 cm from obstacles.
18
Mandatory Nonfunctional Requirements
M.N.1 use Autoware as framework for the software stack.
M.N.2 operate in a geofenced environment.
M.N.3 operate in daylight a well-lit environment.
M.N.4 operate on asphalt.
M.N.5 be powered by an on-board battery.
M.N.6 communicate wirelessly in the ISM band.
M.N.7 have a payload capacity of 1000 kg.
M.N.8 be maintainable: code should be modular and well documented.
M.N.9 be unit testable.
M.N.10 be reliable: subsystem failure modes should be well-defined and detectable.
M.N.11 be compliant with relevant standards and regulations.
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Subsystem Descriptions & Current System Status
35
Sensing Subsystem
36
Sensor Type | Company | Quantity | Purpose |
3D LiDAR | Velodyne VLP-16 | 2 | Localization and P2P navigation for general motion |
RGB Camera | MindVision | 1 | Pod identification, Localization wrt pod, Object detection |
2D LiDAR | Hokuyo UST 20LX | 2 | Localization wrt pod, obstacle detection |
RADAR | Continental SRR308 | 1 | Obstacle detection (to add redundancy to the 2D LiDAR) |
IMU | Xsens MTI-300 | 1 | Localization (augment 3D LiDAR accuracy) |
GPS | SwiftNav Piksi | 1 | Initial pose estimate, Localization (augment 3D LiDAR accuracy) |
Encoder | - | ~2 | Localization (augment 3D LiDAR accuracy) |
Sensing Subsystem - Current Status
Have obtained and interfaced with the 3D LiDARs and Mindvision camera
37
VLP-16 puck mounted on a cart to obtain point cloud map of the NSH B-level
MindVision RGB Camera
Navigation Subsystem
Detection
38
Description Process sensor information (LiDAR, Camera, RADAR) Will detect obstacles using sensor data (2D LiDAR, RADAR and Camera) Will detect AprilTags for Pod Identification using Camera data | Current Status Generated point cloud maps using 3D LiDAR data Used ndt_mapping to generate map of NSH B level AprilTag detection implemented in simulation |
Generated map of our B Level
Pose estimation using AprilTags
39
Navigation Subsystem
Localization
40
Description Localizes and estimates pose in premapped environment Localizes chassis relative to pod when docking/undocking Generates odometry (using IMU and encoder data) | Current Status Implemented using ndt_matching (using 3D LiDAR data) IRL and in simulation Implemented in simulation, relative to pod legs using 2D LiDAR data and relative to AprilTag on Pod Implemented in simulation |
41
Navigation Subsystem
Prediction
42
Description Generates a cost map of system surround- ings using the inputs from the object detector and the precomputed HD vector map. | Current Status Literature review on different methods used in different urban driving scenarios, to draw parallels to our geo-fenced environment needs :
|
Navigation Subsystem
Align
43
Description This block is activated when the system has to dock or undock. Responsible for switching localization input to AprilTags and 2D LiDAR data. | Current Status Fall Semester |
Navigation Subsystem
Mission Planner
44
Description Generates the trajectory waypoints and velocity profile given the cost map and detected landmarks for P2P navigation, and additionally relative pose when planning for approach to pod. | Current Status Implementing in simulation (CARLA) CARLA-Autoware bridge is able to generate waypoints given an initial and final pose for P2P |
45
Navigation Subsystem
Motion Planning
46
Description Generates dynamic trajectory and controls the motion of the system on the basis of the current and desired state. Outputs commanded twist. | Current Status Literature survey of control methods using visual servoing done |
Navigation Subsystem
Actuation
47
Description Performs twist filtering to ensure that constraints are met, and outputs commanded steering and acceleration. | Current Status Implemented in simulation: sending Ackermann steering and velocity commands to vehicle in CARLA simulation |
Docking and Undocking Subsystem
Docking Position Verification
48
Description Once docking position is achieved this node verifies if the correct position has been reached. If yes, the system state is updated to reflect so. If not, then the process of navigating from the PHZ to the pod is repeated until the position is within tolerable limits. | Current Status To be implemented in simulation in Spring. |
Docking and Undocking Subsystem
Adapt
49
Description Once successfully docked or undocked, this block updates the system dynamics and sensor model. | Current Status Fall Semester |
Safety Subsystem
Health Monitoring System
50
Description Monitors functioning and outputs of critical nodes. If failures are detected, this node updates system state to emergency handling state. | Current Status Exploring the use of the Health Checker Function for Autoware |
Safety Subsystem
Emergency Handling
51
Description When the system transitions into the emergency handling state this node ensures:
| Current Status Fall Semester |
Simulation
52
Hardware
53
Project Management
54
Work Breakdown Structure
55
56
57
Schedule
58
Updated Spring Milestones
59
Date | Old Milestones |
1/20 | Autoware Environment Setup |
2/15 | Website Deployed |
2/19 | Simulation Platform Setup |
3/4 | Navigation and Safety Design Ready |
3/25 | Safety Subsystem Ready |
4/8 | Navigation Subsystem Integrated |
4/22 | SVD |
5/4 | CDR |
Date | Milestones |
1/20 | Autoware Environment Setup |
2/15 | Website Deployed |
2/19 | Simulation Platform Setup on CARLA-Autoware |
3/4 | Navigation subsystem design & Simulation setup for docking ready |
3/25 | Navigation subsystem components tested in Simulation: 1. Planning algorithms for docking, 2. Localization of chassis relative to the pod 3. Control algorithms for docking |
4/8 | 1. Navigation subsystem developed 2. Docking subsystem components tested in in Simulation 3. Safety subsystem components Object Detection and Behaviour Prediction developed |
4/15 | Health Monitoring Design Ready, Navigation Subsystem and Docking subsystem Software Integration done. |
4/22 | SVD |
5/4 | CDR |
Fall Milestones
60
Date | Fall Milestone |
31-Aug-2020 | Hardware testing and setup |
15-Sept-2020 | Docking subsystem completed |
30-Sept-2020 | Vehicle models ready for adapt subsystem |
15-Oct-2020 | Safety subsystem completed, adapt subsystem designed |
31-Oct-2020 | All subsystems finished and unit tested |
30-Nov-2020 | Integration testing finished |
High Level Test Plan - Spring
61
Date | Functionality | Test Plan |
Progress Review 3 3/25 | Payload handling zone identification |
|
Planning algorithms for docking in simulation |
| |
Point to Point Navigation in mapped custom CARLA world |
| |
Control algorithms for docking in simulation |
|
62
Date | Functionality | Test Plan |
Progress Review 4 4/8 | Object detection |
|
| Behaviour Prediction |
|
Health Monitoring |
| |
Localization of chassis relative to the pod |
| |
Approach Navigation |
| |
Position Accuracy Verification |
|
Spring Validation Experiments
Equipment: Laptop with Autoware and CARLA, and a precomputed map of the CARLA world.
Test 1: Navigation
63
Test 2: Alignment and Docking
64
Test 3: Safety
65
Test 4: Retracing
66
High Level Test Plan - Fall
67
Date | Functionality | Test Plan |
September | (un)Docking |
|
October | Adapt |
|
November | Safety and Diagnostics |
|
December | Integration Done | Refer to Fall Validation Experiments |
Fall Validation Experiments
Equipment: Chassis, pod, laptop with Autoware and CARLA, and a precomputed map of the CARLA world.
Perform Tests 1 to 4 on actual hardware
Test 5: Lock with Pod and Verify
68
Test 6: Undocking
69
Test 7: Adapt
70
Budget
As we have hardware and sensors being provided by our sponsor or are able to borrow them from inventory, we have only spent some of our budget shown below.
Remaining budget: $4929.55
71
Item | Reason for purchase | Cost |
Adafruit 16-ch 12 bit servo driver (x4) | To control the RC car with a Jetson TX2 while we wait for the chassis from PIX | $70.45 |
Risk Management Table
72
Risk ID | Reqt | Risk | Type | Likelihood | Consequence | Mitigation |
R1 | All MPs | Legacy System might have issues | Technical, Schedule | 3 | 4 | Develop test plan for chassis Test algorithms in simulation |
R2 | All MPs | Delay in arrival of chassis | Schedule, Programmatic | 5 | 5 | Use RC car or cart as replacement Test subsystems in simulation |
R3 | MPs 4, 5, 6, 9, 10 | Sensors stop working | Technical, Cost | 4 | 3 | Sponsor sends sensors early, we perform tests to check usability Use hardware abstraction layer |
R4 | All MPs | Testing infeasible due to weather conditions | Technical, Schedule | 4 | 3 | Plan tests for late spring Test subsystems in NSH B-level |
R5 | All MPs | Communication gap with sponsor | Schedule | 3 | 3 | Document work and provide regular updates |
Risk Management Table
73
Risk ID | Reqt | Risk | Type | Likelihood | Consequence | Mitigation |
R6 | MPs 7, 8, 10 | Issues in Subsystem Integration | Technical, Schedule | 4 | 4 | Perform testing on individual subsystems Develop integration test plan initially. |
R7 | MPs 4, 5, 6 | Overambitious requirements | Technical, Schedule | 4 | 4 | Descope System if Necessary |
R8 | All MPs | Integration issues with code | Technical, Schedule | 2 | 3 | Document work with changes implemented Perform unit testing at each level possible Use a common framework to develop code |
R9 | All MPs | Global pandemic occurs, causing campus shutdown | Programmatic | 1 | 5 | Move everything to simulation Take care of personal health by practicing social distancing and communicating via the internet |
R10 | All MPs | Simulation does not pan out in real-life | Technical | 5 | 4 | Test subsystems individually so that there is no ambiguity in whether the problem is actually due to conversion from sim2real or lies in software |
Risk Likelihood-Consequence Table
74
Consequence
Likelihood
Major Risks:
R1: Legacy system issues
R2: Delay in chassis arrival
R3: Sensors stop working
R4: Testing Infeasible
R6: Integration Issues
R7: Overambitious reqs
R8: Code integration issues
R9: COVID 19
R10: Sim2Real doesn’t work
R1
R2
R3
R4
R5
R6
R7
R8
R9
R10
75
Questions?