Bachelor & PhD Supervision

Academic year 2026–2027 · Individual topics

Specialization: Robotics · Supervisor: Assoc. Prof. Habil. Dr. Eng. Tiberiu-Teodor Cociaș · 18 topics, 26–28 places

Topic 1 - Robotic grasping of unknown objects using 3D shape estimation from stereo images

Two USB cameras reconstruct the object, which is approximated by a geometric primitive (superquadric); a 3D-printed robotic arm computes and executes the grasp.

Mathematics: epipolar geometry, triangulation, Levenberg–Marquardt non-linear optimization, inverse kinematics.
Equipment: 3D-printed servo arm, 2 USB cameras · Estimated budget: €60–80 · Students: 1 student

Topic 2 - Hand-eye camera–robot calibration and evaluation of positioning accuracy

Determining the camera–end-effector transform by solving AX = XB, comparing classical methods and measuring the error on a real arm.

Mathematics: SO(3)/SE(3) groups, quaternions, SVD, least squares.
Equipment: 3D-printed arm, webcam, printed checkerboard · Estimated budget: €50–70 · Students: 1 student

Topic 3 - Image-based visual servoing of a 3D-printed robotic arm

The arm is driven directly from the error measured in the image (IBVS), with a stability analysis of the closed loop.

Mathematics: interaction matrix, Jacobian, Lyapunov functions.
Equipment: 3D-printed arm, webcam · Estimated budget: €50–70 · Students: 1 student

Topic 4 - Learning manipulation tasks from demonstrations (Imitation Learning)

The robot learns a motion from demonstrations (SO-101 leader–follower arm) and generalizes it to new object positions; training runs on Google Colab.

Mathematics: Dynamic Movement Primitives (differential equations), regression, neural networks.
Equipment: open-source SO-101 arm or 3D-printed arm + simulation · Estimated budget: €70–230 · Students: 1 student

Topic 5 - 6D object pose estimation for pick-and-place using neural networks trained on synthetic data

The dataset is generated in Blender; the network detects keypoints from which the object position and orientation are computed and validated with a webcam.

Mathematics: the PnP problem, RANSAC, rotation representations.
Equipment: webcam (optionally a 3D-printed arm) · Estimated budget: €20–70 · Students: 1 student

Topic 6 - Dynamic model identification of a planar robotic arm: analytical model vs. physics-informed neural networks (PINN)

Deriving the Euler–Lagrange model of a 2-DOF arm, identifying its parameters from measurements and comparing it with a hybrid physics + neural network model.

Mathematics: Lagrange equations, linear-in-parameters regression, optimization.
Equipment: 3D-printed planar arm, DC motors with encoders, current sensor, ESP32/STM32 · Estimated budget: ~€50 · Students: 1 student

Topic 7 - Visual-inertial odometry using a smartphone camera and IMU

Estimating the trajectory without GPS by fusing camera-based motion with inertial data; evaluation on a known path.

Mathematics: extended Kalman filter, epipolar geometry, rigid-body kinematics.
Equipment: own smartphone, laptop · Estimated budget: ~€0 · Students: 1 student

Topic 8 - Omnidirectional mecanum-wheel mobile robot: kinematic modelling and trajectory tracking

Building the robot, deriving its kinematic model and implementing a trajectory-tracking controller on a microcontroller.

Mathematics: kinematic Jacobian, non-linear control, tracking-error analysis.
Equipment: 4 mecanum wheels, 4 encoder motors, ESP32/STM32, 3D-printed chassis · Estimated budget: €60–90 · Students: 1 student

Topic 9 - Lateral control of a small-scale autonomous vehicle based on lane detection

Lane detection with a neural network and comparison of Pure Pursuit, Stanley and MPC controllers on a steered vehicle.

Mathematics: bicycle model, homography, model predictive control (MPC).
Equipment: 3D-printed chassis with steering servo, ESP32-CAM, laptop · Estimated budget: €50–80 · Students: 1 student

Topic 10 - Semantic occupancy maps for mobile robot navigation using neural segmentation on an embedded platform

Real-time image segmentation with a quantized network, projected into a probabilistic occupancy grid.

Mathematics: Bayesian log-odds update, perspective projection, network quantization.
Equipment: Raspberry Pi + camera (or laptop processing) · Estimated budget: €20–120 · Students: 1 student

Topic 11 - Gesture-based teleoperation of a robotic arm using hand pose estimation

The hand skeleton detected by the camera is mapped into the robot workspace, with tremor filtering.

Mathematics: forward and inverse kinematics, coordinate transforms, Kalman filter.
Equipment: 3D-printed arm, webcam · Estimated budget: €40–60 · Students: 1 student

Topic 12 - 3D scanner based on a 1D lidar mounted on a pan-tilt mechanism

The lidar is rotated on two axes to obtain a point cloud of a room; scans from several positions are registered into a single model.

Mathematics: spherical coordinates, measurement error modelling, ICP, SVD.
Equipment: 1D lidar (available), 2 servos, 3D-printed mount · Estimated budget: ~€15 · Students: 1 student

Topic 13 - Impedance control of a 2-DOF robotic arm driven by BLDC motors

The arm behaves as a mass–spring–damper system on contact with a human; torque is estimated from motor current (SimpleFOC).

Mathematics: manipulator dynamics, second-order differential equations, stability.
Equipment: BLDC gimbal motors, AS5600 magnetic encoders, FOC driver, 3D-printed parts · Estimated budget: €70–100 · Students: 1 student

Topic 14 - Fault and anomaly detection on a robotic axis using motor signals and neural networks

Faults (imbalance, friction, backlash) are induced on a 3D-printed axis and detected from current and vibration with an autoencoder.

Mathematics: FFT, spectral analysis, Mahalanobis distance, neural networks.
Equipment: DC motor, INA219 current sensor, MPU6050 accelerometer, ESP32 · Estimated budget: €30–40 · Students: 1 student

Topic 15 - 2D mapping with a rotating 1D lidar mounted on a mobile mini-robot

A motor with encoder spins the lidar to obtain a 2D scan, and an occupancy map is built from one of the research institute's mini-robots.

Mathematics: polar coordinates, motion-distortion correction, scan matching, Bayesian log-odds update.
Equipment: 1D lidar and mini-robot (available), encoder motor · Estimated budget: €10–15 · Students: 1 student

Academic year 2026–2027 · Collaborative platforms

Several students work on the same platform, but each owns a distinct module with its own mathematical model, its own evaluation metrics and a documented interface to the other modules.

Topic 16 - Robotic sorting cell with a delta robot and conveyor belt (4–5 students)

A 3D-printed delta robot with a suction gripper picks parts from a moving conveyor and sorts them by category. Modules communicate through interfaces agreed from the start (ROS2/MQTT).

Modules (one student each):

1. Delta robot modelling and control - analytical forward/inverse kinematics, Jacobian, singularities, workspace
2. Part detection and classification with computer vision - pinhole camera model, conveyor-plane homography, precision/recall metrics
3. 1D lidar profilometry above the conveyor - 3D profile from lidar + encoder, height and volume estimation, Bayesian fusion with the camera
4. Conveyor tracking and interception trajectory planning - Kalman filter, 5th-order polynomial trajectories, cycle-time optimization
5. (optional) Conveyor drive and cell coordination - DC motor model, PID tuning, state machine, performance indicators

Equipment: 3 servos/steppers, vacuum pump with suction cup, encoder motor, camera, microcontroller, 3D-printed parts; 1D lidar available · Estimated budget: €100–150 total

Topic 17 - Autonomous mobile robot for intralogistics (4 students)

A differential-drive robot that transports parts in a lab or warehouse; each student develops a distinct module.

Modules (one student each):

1. Motion modelling and control - kinematic and dynamic model, MPC/PID control on STM32
2. Localization and mapping (SLAM) - particle filter, probabilistic sensor models
3. Path planning - A* / D* Lite, Dynamic Window Approach for dynamic obstacles
4. Neural perception - detection of stations, ArUco markers and people, relative pose estimation

Equipment: 3D-printed chassis, encoder motors, low-cost LIDAR (LD06/LD19), Raspberry Pi, ESP32 · Estimated budget: €180–250 total

Topic 18 - Fleet of coordinated mobile mini-robots (3–4 students)

Several of the research institute's mini-robots work together in an arena, localized by a ceiling-mounted camera.

Modules (one student each):

1. Global localization with camera and markers - camera calibration, homography
2. Formation control with consensus algorithms - graph Laplacian, stability analysis
3. Task allocation and multi-robot planning - Hungarian algorithm, path-conflict resolution
4. Real-time firmware and communication - protocols, latency, synchronization

Equipment: mini-robots available at the institute, webcam · Estimated budget: ~€20 total