Fixing navigation errors when a warehouse robot lacks boundary information
Case context: You are building an autonomous warehouse robot. It moves poorly in narrow aisles because the route-planning step needs aisle-edge locations, but the current pipeline does not provide them.
Question: According to good ML system design, what change should you make to the pipeline, and why is it better than forcing an existing module to do this job too?
Sample answer: Add a dedicated module that detects aisle edges and sends that information to the route-planning step. This is better because it supplies the missing signal directly and keeps the other modules simpler to design and train.
Key points:
- Add a dedicated aisle-edge detector.
- Pass the needed boundary information to route planning.
- Avoid overloading existing modules with extra complexity.
Rubric: The response should identify adding a new boundary-detection module and explain that it gives the planner the missing information while avoiding unnecessary complexity in other parts of the pipeline.
0
1
Tags
Machine Learning
Deep Learning
Supervised Learning
Dive into Deep Learning @ D2L
Data Science
Machine Learning Strategy
Machine Learning Yearning @ DeepLearning.AI
Related
Keeping a Downstream Module Focused During a Redesign
Why would a navigation team add a separate curb-detection module to its driving stack?
True or False: A lane-marking detector can be added so the planner receives lane-position information it did not already have.
When a warehouse robot's route planner is missing shelf-location information, Ng recommends redesigning it by adding a _____ component.
Why add a curb-detection module to an autonomous wheelchair navigation pipeline?
If a factory robot also needs aisle-label information to navigate safely, that is a sign the control pipeline should be redesigned to include an aisle-label detector.
A shelf-edge detector gives the route planner important, previously missing information about _____.
Match each redesign element to its role when adding a road-edge detector.
Order the reasoning steps for adding an aisle-marker detector to a warehouse robot pipeline.
Why is a separate pothole detector often better than folding pothole detection into an existing road-condition module?
The route-planning module would perform just as well whether or not it receives lane-position information.
Using a separate road-sign classifier can keep one part from becoming overly _____ to design or train.
Match each design choice to the result it produces in a robot navigation redesign.
Order the steps for revising a self-driving shuttle pipeline to include a lane-marking detector.
Adding a Lane-Line Detector to a Driving Stack
Fixing navigation errors when a warehouse robot lacks boundary information
What downstream module uses the new detector output, and why?