knaxuslearn · build · ship

Roadmaps / Robotics / Interview topics

Reference

Interview topics

Use these for robotics, autonomy and embedded roles. Each follow-up is the question a good interviewer asks next. The Staff line is what lifts the answer.

Phase 1

QuestionFollow-upA Staff answer adds
Topic, service or action: how do you choose?What does QoS change?A rule for each message type, tested for dropped data
Explain the transform tree of a mobile robotWhy are map to odom and odom to base_link separate?What breaks if a node uses the wrong one, and a check that catches it
What does use_sim_time do?What goes wrong when it is mixed?A launch convention that sets it once for the whole system
How does ros2_control separate controllers from hardware?Why does that help?Swapping simulation and real hardware without changing controller code
How do you measure latency in a robot pipeline?Where does it come from?Percentiles, not averages, and a budget per stage

System design prompt (20 minutes): the software architecture of a delivery robot. Outline: requirements and numbers (3 min), nodes and topics (5), frames and time (4), simulation and real hardware (4), testing (4).

Phase 2

QuestionFollow-upA Staff answer adds
Why does wheel odometry drift?How do you reduce it?Which sensor fixes which error, and the one drift no sensor removes
Explain a Kalman filter without equationsWhat does the covariance do?How you would detect a filter that is overconfident
How do you tune sensor fusion?How do you know it improved?A repeatable path, a metric and a table before and after
Where does lidar SLAM fail?How would you detect it online?A health signal and a fallback, not only a better algorithm
How do you debug a robot problem you cannot repeat?What do you record?Bags with timestamps and a replay harness, set up before the failure

System design prompt (20 minutes): localisation for a robot in a changing warehouse. Outline: sensors and failure modes (4 min), estimator design (5), map strategy (4), health monitoring (4), testing (3).

Phase 3

QuestionFollow-upA Staff answer adds
How does the Nav2 stack fit together?Where does it fail most often?A failure taxonomy with rates, from your own runs
AMCL or SLAM for a deployed robot?What if the map changes?When to remap, and who owns the map
Which local controller would you choose?What does it cost in tuning?A comparison on the same runs, and the situation where the choice flips
What is a costmap inflation radius?How do you pick it?Clearance against passability, measured in the narrowest aisle
How do you report navigation performance?Why does the trial count matter?Counts with intervals, failure classes and a rerun on the same seeds

System design prompt (20 minutes): a fleet of 20 delivery robots sharing aisles. Outline: requirements (3 min), single-robot autonomy (4), traffic and deadlocks (5), monitoring and intervention (4), rollout (4).

Phase 4

QuestionFollow-upA Staff answer adds
How do you make a mobile robot fail safely?What if the computer hangs?An independent hardware stop and a stated residual risk
How do you size a stop zone?What changes at higher speed?Reaction plus braking distance, and a margin you can defend
How do you test a robot in CI?What can simulation not tell you?Seeded faults that must fail the job, and the list of what only hardware shows
What causes jitter in a ROS 2 control loop?How do you find it?Tracing, executor and callback group choices, and a measured budget
How would you roll out a software update to robots?What if it fails in the field?Staged rollout, rollback, and a metric that triggers it

System design prompt (20 minutes): a safety and reliability plan for robots working near people. Outline: hazards and standards you must check (4 min), controls and testing (5), monitoring (4), incident response (4), evidence for a review (3).