Q: Explain what forward kinematics and inverse kinematics are, and what method did you use to solve IK for a 6-DOF robot arm in a real project?

Clear definitions: FK - given joint angles, compute end-effector position and orientation (unique solution); IK - the reverse: given desired end-effector position, find joint angles (may have multiple solutions or no solution). IK methods: (1) Analytical - closed-form solution, fast but only usable for robots with specific structure (wrist singularity); (2) Numerical - Jacobian pseudo-inverse, iterative, more general but slower and may not converge; (3) KDL (Kinematics and Dynamics Library) in ROS; (4) MoveIt2 with IKFast or TRAC-IK. Singularities are an important practical point - ask about detecting and handling singularities when robot joints align. Strong candidates know why TRAC-IK is often preferred over the default KDL solver in MoveIt2.

Q: When deploying a collaborative robot (cobot) in a production line in Vietnam, what safety standards must you comply with and what is the risk assessment process?

Applicable international standards: (1) ISO 10218-1/-2: Robots and robot systems - safety requirements; (2) ISO/TS 15066: Collaborative robot systems - speed & separation monitoring, power & force limiting; (3) IEC 62061 or ISO 13849 for safety functions. Process: (1) Hazard identification - collaborative workspace, pinch points; (2) Risk estimation - severity × probability × avoidability; (3) Risk reduction: inherently safe design → safeguarding → information for use; (4) Validation: force measurement, speed measurement, CE marking for EU export. Vietnam reality: QCVN has no dedicated standard, factories typically follow ISO or FDI parent company standards. This is an especially important question as cobots are increasingly common in Vietnamese factories (Universal Robots, FANUC CRX, ABB GoFa). Good candidates will know the 4 collaborative operation modes in ISO/TS 15066 and when to apply each.

Q: Where do you think the robotics industry in Vietnam currently stands and what are the opportunities and challenges for robotics engineers in the next 5 years?

Realistic assessment: (1) Current state - industrial robots mainly in FDI factories (automotive Vinfast/Thaco, electronics Samsung/LG/Intel), Vietnamese SMEs use very few due to cost; (2) 5-year trends: robot costs dropping sharply (Universal Robots cobots under $30K), logistics automation (Giao Hang Nhanh, Lazada warehouse), agriculture (spraying drones, harvesting); (3) Opportunities: system integration, robotics application engineering, AI/vision for robots; (4) Challenges: shortage of properly trained robotics talent, high upfront cost, local maintenance capability. This question assesses the candidate's market awareness and strategic thinking. Strong candidates will have specific opinions based on real experience, not generic talk about 'the AI and robot future'. Ask further about their own development plans.

Question 1

How do you understand ROS2 architecture? Explain the differences from ROS1 and why those changes matter for robots in industrial production environments.

Accepted Answer

Key differences: (1) ROS1 uses ROS Master (single point of failure) - ROS2 uses DDS (Data Distribution Service) for peer-to-peer discovery, no master; (2) Real-time support: ROS2 supports real-time OS (rclcpp with real-time executor), ROS1 does not; (3) Security: ROS2 has SROS2 with DDS security plugins, ROS1 does not; (4) Multi-robot: native multi-robot coordination via DDS; (5) QoS policies to control communication reliability and latency. Industrial relevance: real-time and reliability are mandatory requirements. Probe further on specific experience with DDS vendors (FastDDS, CycloneDDS) and how to configure QoS for real-time applications. Candidates who work with industrial robots will know about Nav2 (navigation) and MoveIt2 (manipulation).

Question 2

Have you performed sensor fusion? Explain how you fused data from an IMU and camera (or LiDAR) for more accurate pose estimation, and which algorithms you used.

Accepted Answer

Explain by pipeline: (1) Why fusion is needed - IMU is good for high-frequency short-term (drifts over time), camera/LiDAR is good for low-frequency absolute positioning; (2) Extended Kalman Filter (EKF) - linearizes non-linear systems, used for Visual-Inertial Odometry (VIO); (3) Tightly-coupled vs. loosely-coupled fusion; (4) In practice: VINS-Mono or VINS-Fusion for visual-inertial, LOAM or LIO-SAM for LiDAR-IMU; (5) Calibration: temporal calibration (time offset between sensors), spatial calibration (extrinsic matrix). Calibration between sensors is often overlooked but is the deciding factor for accuracy. Ask about experience with Kalibr or camera-imu calibration tools. Strong candidates will discuss uncertainty propagation.

Question 3

What have you used computer vision for in robotics? Describe specifically how you detected and estimated the pose of an object for robot grasping.

Accepted Answer

6-DOF pose estimation pipeline: (1) Detection - YOLOv8 or SSD to locate object in image; (2) Pose estimation methods: PnP algorithm if 3D model + 2D keypoints are known; depth camera (RealSense, ZED) to get pointcloud; FoundationPose or ORB-SLAM3 for markerless; (3) Coordinate transformation - camera frame to robot base frame using hand-eye calibration (eye-in-hand vs. eye-to-hand); (4) Grasp planning - GraspNet or geometric approach based on object shape; (5) Execution - MoveIt2 to plan and execute trajectory. Hand-eye calibration is a step that is frequently underestimated but determines overall accuracy. Ask about experience with calibration tools (easy_handeye ROS package) and how to handle calibration when the robot vibrates or the tool changes.

Question 4

Have you worked with embedded systems for robotics? Compare when to use an RTOS (FreeRTOS, Zephyr) vs. standard Linux for robot control tasks.

Accepted Answer

Decision based on timing requirements: Use RTOS when: hard real-time needed (< 1ms latency, deterministic), e.g. motor control loops (PID at 10kHz), safety-critical interrupt handling, low-power microcontrollers (STM32, ESP32); Use Linux (RT-preempt patch) when: soft real-time needed (< 10ms), running ROS2, processing vision/sensor fusion, network connectivity. Practical architecture: MCU (STM32) running RTOS for low-level control ↔ SBC (Jetson, RPi) running Linux for high-level logic, communicating via micro-ROS or CAN bus. micro-ROS is an important bridge between microcontrollers and the ROS2 ecosystem. Candidates doing real robotics work in Vietnam often work with ESP32 + micro-ROS or STM32 + FreeRTOS. Ask about experience with CAN bus or EtherCAT for motor control.

Question 5

Explain what forward kinematics and inverse kinematics are, and what method did you use to solve IK for a 6-DOF robot arm in a real project?

Accepted Answer

Clear definitions: FK - given joint angles, compute end-effector position and orientation (unique solution); IK - the reverse: given desired end-effector position, find joint angles (may have multiple solutions or no solution). IK methods: (1) Analytical - closed-form solution, fast but only usable for robots with specific structure (wrist singularity); (2) Numerical - Jacobian pseudo-inverse, iterative, more general but slower and may not converge; (3) KDL (Kinematics and Dynamics Library) in ROS; (4) MoveIt2 with IKFast or TRAC-IK. Singularities are an important practical point - ask about detecting and handling singularities when robot joints align. Strong candidates know why TRAC-IK is often preferred over the default KDL solver in MoveIt2.

Question 6

When deploying a collaborative robot (cobot) in a production line in Vietnam, what safety standards must you comply with and what is the risk assessment process?

Accepted Answer

Applicable international standards: (1) ISO 10218-1/-2: Robots and robot systems - safety requirements; (2) ISO/TS 15066: Collaborative robot systems - speed & separation monitoring, power & force limiting; (3) IEC 62061 or ISO 13849 for safety functions. Process: (1) Hazard identification - collaborative workspace, pinch points; (2) Risk estimation - severity × probability × avoidability; (3) Risk reduction: inherently safe design → safeguarding → information for use; (4) Validation: force measurement, speed measurement, CE marking for EU export. Vietnam reality: QCVN has no dedicated standard, factories typically follow ISO or FDI parent company standards. This is an especially important question as cobots are increasingly common in Vietnamese factories (Universal Robots, FANUC CRX, ABB GoFa). Good candidates will know the 4 collaborative operation modes in ISO/TS 15066 and when to apply each.

Question 7

Where do you think the robotics industry in Vietnam currently stands and what are the opportunities and challenges for robotics engineers in the next 5 years?

Accepted Answer

Realistic assessment: (1) Current state - industrial robots mainly in FDI factories (automotive Vinfast/Thaco, electronics Samsung/LG/Intel), Vietnamese SMEs use very few due to cost; (2) 5-year trends: robot costs dropping sharply (Universal Robots cobots under $30K), logistics automation (Giao Hang Nhanh, Lazada warehouse), agriculture (spraying drones, harvesting); (3) Opportunities: system integration, robotics application engineering, AI/vision for robots; (4) Challenges: shortage of properly trained robotics talent, high upfront cost, local maintenance capability. This question assesses the candidate's market awareness and strategic thinking. Strong candidates will have specific opinions based on real experience, not generic talk about 'the AI and robot future'. Ask further about their own development plans.