test-algo-on-HalfCheetah.mp4
bench-inspection-demo.mp4
navi-cluttered-env.mp4
Immediately after the submission of the thesis, someone released a same project and the performance is far better than mine.
| Supervisor | Assessor | Mark | |
|---|---|---|---|
| Specification | N/A | 60 | 60 (15%) |
| Interim Presentation | 71 | 66 (Dr Alex Phillips stood in for Prof Jason Ralph) | 69 (15%) |
| Bench Inspection | 79 | 73 | 76 (15%) |
| Thesis (with one day late penalty) | 85 - 5 | 81 - 5 | 78 (55%) |
| Overall | N/A | N/A | 74 |
gantt
dateFormat YYYY-MM-DD
Literature review: literature_review, 2022-10-1, 90d
Interim presentation: milestone, 2022-12-13,
section Preparation
Implement SOTA algos: 2022-10-24, 50d
Sanity test the algos on popular environments: test_algo, 2022-11-15, 28d
Algorithms were ready : milestone, 2022-12-13, 0d
Spawn a robot into a simulator: load_robot, after test_algo, 3d
Program the training env: program_env, after load_robot, 37d
Match the interface between algo and env: match_interface,after program_env, 7d
Training in an empty env: traning_in_empty_env, after match_interface, 9d
Trained a workable policy in an empty env: milestone, 2023-2-7,
section First Attempt
Set up an indoor environment filled with obstacles for training: setup_indoor_env, after traning_in_empty_env, 7d
Tinker the reward function: after setup_indoor_env, 10d
Training in a cluttered environment: training_in_cluttered_env, after setup_indoor_env, 14d
Observable intelligent mapless navigation behaviour: milestone, after training_in_cluttered_env
section Training-Evaluation Loop
Set up another environment for evaluating the performance of trained policy: setup_devel_env, after training_in_cluttered_env, 7d
Repeat training and evaluating: train_eval, after setup_devel_env, 2023-03-23
Bench inspection: milestone, after train_eval
section Bachelor's Thesis
Collect figures: 2023-03-23, 8d
Compose: 2023-03-23, 2023-04-18
Edit and proofread: 2023-04-18, 2023-04-20
Submitted: milestone, 2023-04-20