Skip to content
Tech Interview Prep
Technical interview guide

Robot Kinematics

The geometry of robot motion: translating between joint angles and the position of the robot's end effector.

Read
45 min
Practice MCQs
25
Interview Q&A
25
Edition
v4
Editorial status
Reviewed
Relevant for
Robotics Engineer

Scope: Modern Robotics, ROS 2 Rolling tf2/URDF, MoveIt 2, Drake, Pinocchio, MATLAB Robotics System Toolbox, SciPy, and Eigen documentation reviewed 2026-09-04.

Curated: · Written: · Reviewed:

25 questions

MCQ-1A planar two-link arm has link lengths L1 = 1.0 m and L2 = 1.0 m. The shoulder joint sits at θ1 = 30° from the base x-axis, and the elbow joint is at θ2 = 60° measured relative to the extension of link 1. Using forward kinematics, what is the end-effector position (x, y) in the base frame?
MCQ-3You are adding a validation check to a robot controller. A third-party kinematics library returns this 3×3 matrix as the end-effector orientation: M = [[0, 1, 0], [1, 0, 0], [0, 0, 1]]. Before using M to transform poses, you check whether it is a valid rotation. What is the correct conclusion?
MCQ-6A robot has two rigid transforms: T_A_B maps points from frame B to frame A, and T_C_D maps points from frame D to frame C. A planner multiplies them as T_A_B * T_C_D and claims the result maps D to A. Which condition makes that claim correct?
MCQ-2A robot arm's controller publishes the pose of its end-effector as a 4x4 homogeneous matrix T_base_tool. A vision node needs to convert a point it measures in the tool frame into the base frame, and it currently computes p_base = T_base_tool * p_tool. A second node instead needs the tool-frame coordinates of a point given in the base frame, and it reuses the same matrix by computing p_tool = T_base_tool * p_base. The tool pose is a pure rotation of 90 degrees about the base z-axis with the tool origin at (1, 0, 0) in base coordinates. For a base-frame point p_base = (1, 1, 0), what does the second node's computation p_tool = T_base_tool * p_base actually produce, and why is it wrong?
MCQ-17A 6-axis arm reports its flange (tool0) pose in the base frame as position (500, 200, 800) mm, with the flange frame rotated +90° about the base Z axis (flange +X points along base +Y, flange +Z along base +Z). A gripper is mounted with its TCP at (150, 0, 0) mm in the flange frame. Where is the TCP in the base frame?
MCQ-5A robot arm interpolates its tool orientation with SLERP between q_start = (1, 0, 0, 0) and q_goal = (−0.7071, 0, 0, 0.7071). The controller applies the standard sign check before interpolating. What is the tool orientation at t = 0.25?
MCQ-4A 6-axis robot wrist is commanded with ZYZ Euler angles (phi, theta, psi), each angle limited to [-180 deg, +180 deg]. The controller linearly interpolates the three angles between consecutive commands over 100 ms. Command A is (phi, theta, psi) = (40 deg, 0 deg, 10 deg) and command B is (30 deg, 0 deg, 40 deg). The base frame and tool frame are unchanged, and no other compensation is active. At 50 ms into the move, what is the tool's rotation about Z?
MCQ-18A warehouse robot carries a lidar on a turret that yaws at 90°/s. The TF tree is map → base_link → turret_link → lidar_link, and the turret joint is published at 20 Hz. A perception node needs the lidar pose in map at the instant a scan was captured (scan_stamp). It computes T1 = lookupTransform("map", "base_link", scan_stamp) and T2 = lookupTransform("base_link", "lidar_link", ros::Time(0)), then pose = T1 * T2. The newest turret sample is up to 50 ms newer than scan_stamp, and the lidar sits 0.8 m from the turret axis. What is the resulting pose error, and what should the node do?
MCQ-19A 6-DOF arm is at configuration q, where the end-effector frame is R(q) = Rot_z(90 degrees). With the other joints held fixed, the geometric Jacobian's first column predicts an angular velocity of [0, 0, 1] rad/s when joint 1 moves at 1 rad/s. A teammate checks that column by computing D = (R(q + h e1) - R(q))/h with h = 1e-6 s, then reading the angular part as [D(3,2), D(1,3), D(2,1)]^T. That gives [0, 0, 0], so the teammate says the Jacobian is wrong. Which conclusion is correct?
MCQ-7A 6-DOF arm is at a configuration where the geometric Jacobian J (6x6, expressed in the base frame) has det(J) = 0. The controller commands joint rates qdot = [0.1, 0, 0, 0, 0, 0] rad/s and the arm does not move. Which statement explains this?
MCQ-14A 7-DOF arm tracks a 6-DOF end-effector pose with a resolved-rate controller, and at the current configuration the 6×7 task Jacobian J has full row rank. Which joint velocity is a null-space motion?
MCQ-8A 2-link planar arm with L1 = L2 = 0.5 m sits at θ1 = 0°, θ2 = 0°, fully extended along +x. A resolved-rate controller using the Moore–Penrose pseudoinverse is commanded to move the tip at v = (0.1, 0) m/s. What joint rates does it command?
MCQ-9A 2-link planar arm with L1 = L2 = 0.5 m is at θ1 = 0°, θ2 = 170°. Its Jacobian has smallest singular value σ_min ≈ 0.087 m/rad. What does this value tell the controller?
MCQ-24A 6-DOF arm with no redundant joints tracks a Cartesian path through a kinematic singularity. At the configuration where the Jacobian's smallest singular value is sigma_min = 0.005, a commanded end-effector twist of 0.05 m/s along the weak task direction produces a Moore-Penrose pseudoinverse joint command of 10 rad/s, above the 2 rad/s drive limit. Which modification of the differential inverse-kinematics formulation bounds the joint command while accepting a controlled tracking error along the weak direction?
MCQ-10A planar two-link arm has equal links L1 = L2 = 0.4 m. Joint 2 (the elbow) is limited to 0°–180°, so it cannot bend the other way. The controller must drive the tool tip to (0.4, 0.4) m. Which joint-angle pair (θ1, θ2) reaches the target without violating the elbow limit?
MCQ-12A 2-link planar arm (L1 = L2 = 0.5 m) runs damped least-squares IK with no random restart. The controller seeds the solver at θ1 = 0°, θ2 = 0° and asks the tip to reach (0.6, 0) m. What does the solver return?
MCQ-11A 6-DOF arm with a 1.20 m maximum reach runs a damped-least-squares IK solver whose convergence test is 'joint update norm < 1e-4 rad'. Commanded to a target 1.45 m from the base, it returns converged=true after 42 iterations with a final Cartesian error of 0.25 m. What is the correct diagnosis?
MCQ-16A 2-link planar arm with 0.5 m links runs inverse kinematics at 200 Hz with no limits inside the solver; the loop clips the returned angles to joint 1 in [-150, 150] deg and joint 2 in [-30, 60] deg before commanding the servos. For a test target at (0.707, 0) m the solver returns q1 = -45 deg, q2 = +90 deg, and the arm stops 259 mm from the target. The offline planner solves the same target with the joint-2 bound inside the optimization and returns the feasible configuration that minimizes the end effector's distance to the target. What does that planner return, and how far off is the end effector?
MCQ-13A 6-DOF arm runs damped least-squares IK at 1 kHz. The task residual stacks a position error in metres and an orientation error in radians, and the solver minimises the plain 2-norm of that stacked vector. On a Cartesian move the tool tracks position to 0.4 mm but orientation drifts 3.5° off target, and the joint velocities stay well below their limits. What is the most likely cause of the orientation drift?
MCQ-15A 6-DOF arm follows a Cartesian path at 500 Hz (2 ms per cycle). Each cycle the controller calls a numerical pose IK solver seeded with the previous cycle's joint solution; it converges in about 1 ms. Along one segment the wrist passes near a singularity and the returned joint solution flips by roughly 40° between consecutive cycles, tripping the joint-velocity limit. Which change addresses the cause of the fault?
MCQ-20A 6-axis arm with a 0.9 m reach is commanded to move its tool from a pose in free space to a target 0.7 m from the base, inside a fixture. The IK solver returns a configuration with position residual 0.3 mm and orientation residual 0.05°, both inside the controller's 1.0 mm / 0.5° tolerance; the pose has two IK branches. The controller interpolates in joint space from the current configuration to that solution over 12 waypoints. At waypoint 7, the elbow link's clearance to the fixture is -18 mm. What is the correct next step?
MCQ-25A 6-DOF arm's inverse kinematics is validated by sampling 5,000 random joint configurations, running forward kinematics to get each TCP pose, feeding that pose back through IK, and checking the returned joint vector matches the original to within 1e-6 rad. All 5,000 pass. On the floor the arm places parts 12 mm off target at full extension but within 0.5 mm near the base. What does the passing test fail to establish?
MCQ-22A 6-axis arm is calibrated by least-squares identification of 24 geometric parameters (link lengths, joint offsets, twists) from 60 measured poses. In all 60, joints 2, 3 and 5 sit between 88 and 92 degrees, while joints 1, 4 and 6 travel through most of their ranges. The fit residual drops from 1.7 mm to 0.2 mm, but 40 held-out poses spread over the whole joint space average 2.6 mm of error. What should the team do next?
MCQ-23A 6-DOF arm is commanded to move quasi-statically (negligible joint acceleration) to a target pose. The inverse-kinematics solver returns a solution with all six joints inside their limits and no self-collision, and the end-effector sits 0.9 m from the shoulder axis. With a 5 kg payload in the gripper, the shoulder joint faults on torque saturation and the arm stops short of the target. Which conclusion fits these observations?
MCQ-21A warehouse robot uses a differential drive with wheel radius r = 0.1 m and wheel separation L = 0.5 m. At pose (x = 0, y = 0, θ = 0) it receives the body-frame twist command (v_x = 0, v_y = 0.5 m/s, ω = 0) to slide sideways into a dock. The controller applies the standard differential-drive inverse kinematics. What does the controller do with this command?