# Week 8 — Robust obstacle-avoiding rover capstone

DIS OpenLab · Foundations of Autonomous Systems · v0.1.0-prepilot · 2026-10-15
Track: simulation-first capstone; optional approved hardware extension
License proposal: CC BY 4.0 for original text, pending rights review

## Capstone brief

Create a reproducible rover controller that progresses through a declared course, responds to an obstacle, reaches a safe terminal condition, and remains understandable under two named disturbances. Submit enough evidence for another team to repeat the work and identify its limitations.

This material does not include a verified reference outcome. The supplied controller and world are starting points, not proof that a completed mission exists.

## Learning objectives and required outcomes

Learners must demonstrate:

- a system and interface model;
- bounded controller logic with invalid/stale/disabled safe states;
- kinematic predictions with units and frames;
- timestamped measurements and commands;
- at least two repeatable disturbance scenarios;
- fault investigation and a justified change;
- reproducible configuration, source, and run instructions;
- honest separation of computational, native Webots, and physical evidence.

## Safety and authorization gate

Simulation may proceed under normal computer-use rules. Hardware is optional and may run only after host authorization, adult supervision, approved low-voltage equipment, barriers, manual isolation, preflight, and a documented stop condition. A failed safeguard or critical equipment fault ends the run.

## Protocol template

1. **Question:** What behavior will be demonstrated and under which disturbances?
2. **Configuration:** source version/hash, parameters, world/arena, geometry, runtime, OS, simulator version, and evidence destination.
3. **Start state:** pose, controller state, obstacle state, and enabled state.
4. **Success criteria:** declared before execution; include safe terminal behavior.
5. **Trials:** at least one nominal and two disturbed scenarios, repeated where resources permit.
6. **Metrics:** end pose/error, completion, stop reason, time, minimum clearance, saturation/fault count.
7. **Failure response:** preserve logs, stop safely, describe symptoms, form competing hypotheses, test one change.
8. **Report:** results, negative results, limitations, reproduction steps, attribution, and next test.

## Suggested disturbances

Choose two that can be controlled and documented:

- effective wheel mismatch;
- sensor noise or bias;
- changed obstacle position;
- bounded command limit;
- one delayed update that should trigger watchdog stop.

Do not introduce random variation without saving the seed and generation method.

## Evidence register

| Artifact | Filename/link | Version/hash | Evidence class | Reviewer result |
|---|---|---|---|---|
| Controller source | | | computational / Webots / hardware | |
| Configuration | | | | |
| Nominal telemetry | | | | |
| Disturbance A telemetry | | | | |
| Disturbance B telemetry | | | | |
| Test output | | | | |
| Screenshot/photo (if permitted) | | | | |
| Engineering report | | | | |

## Engineering report structure

1. Abstract: what was attempted, without inflated claims.
2. System: frames, interfaces, assumptions, safety boundary.
3. Method: exact protocol and predeclared criteria.
4. Results: tables/plots with units and evidence class.
5. Debugging: failure, competing causes, test, and change.
6. Limitations: what was not measured or generalized.
7. Reproduction: clean commands, versions, configuration, files.
8. Attribution and license status.

## Assessment rubric (100 points)

| Dimension | Points | Minimum evidence |
|---|---:|---|
| Model understanding | 15 | correct system/mathematical explanation |
| Working controller | 25 | repeatable required scenarios |
| Measurements/testing | 20 | timestamps, units, comparable runs |
| Engineering debugging | 15 | fault analysis and responsible mitigation |
| Documentation/reproducibility | 15 | source, versions, configuration, commands |
| Safety/ethics/collaboration | 10 | completed gates and appropriate attribution |

A suggested passing demonstration is 70/100 with no unresolved critical safety concern. It is not a high-stakes credential or professional qualification.

## Instructor review checklist

- [ ] Claims match the evidence class.
- [ ] No physical result is inferred from simulation.
- [ ] Inputs, units, frames, timestamps, and versions are present.
- [ ] The run can be repeated without unpublished instructions.
- [ ] Negative/failure evidence is retained.
- [ ] Safety and privacy restrictions were followed.
- [ ] Third-party assets and sources are attributed.

## Low-bandwidth and accessibility option

A learner may submit code on removable media, printed telemetry excerpts, handwritten calculations, and a text-only report. Simulation-only completion is fully valid. Assign roles across coding, testing, safety review, measurement, and documentation; no learner must operate physical hardware to demonstrate the core outcomes.

## Final reflection

State the strongest claim your evidence supports, the strongest claim it does **not** support, and the next experiment that should occur before anyone relies on the controller in a physical setting.
