Ken KADILAR

An autonomous, biosignal-driven sorting cell built around the leadscrew gripper from my M.Sc. thesis. An overhead camera classifies each object by colour, the gripper grasps it at a muscle-taught force under closed-loop PID control, and a rotation arm sorts it to the matching bin. A soft PLC (OpenPLC, IEC 61131 in both Structured Text and Ladder) runs the 18-state cycle and a latching E-stop, and Modbus TCP ties one master to three independent slaves. A design-of-experiments run characterizes grip force against object damage.

• OpenPLC / IEC 61131 (ST + Ladder)
• Modbus TCP (1 master, 3 slaves)
• closed-loop current PID
• OpenCV colour classifier
• EMG grip-force teach
• 2x5 factorial DOE (n=50)

A standalone myoelectric gripper on real STM32 firmware: a forearm EMG signal is classified on-chip at 1 kHz, the gripper actuates, and the board broadcasts its state over a CAN bus, with a hardware watchdog and a signal-loss fail-safe. It rebuilds my M.Sc. thesis on an industrial real-time stack: FreeRTOS tasks, timer + DMA sensing, and a hand-written, register-level MCP2515 CAN driver (no vendor library), verified two-node against a separate CAN node. The gripper is a geared mechanism I designed in SolidWorks (tooth profiles from a gear-design tool I wrote), so the project spans mechanism, electronics, and firmware.

• STM32F446 / C++ (PlatformIO)
• FreeRTOS (4 tasks)
• CAN bus (MCP2515)
• hand-written driver
• 1 kHz timer + DMA
• IWDG watchdog + fail-safe
• SolidWorks geared gripper

A multi-grip prosthetic hand driven by a single surface-EMG channel. Two muscle impulses are classified on-device with Dynamic Time Warping on ESP32 and combined into a 2-bit code that selects one of four grips, at a measured 92% pattern accuracy and 414-644 ms impulse-to-motor latency on a 3D-printed PLA mechanism (DC motor + worm gear + encoder). Built full-stack: a hand-fabricated double-sided KiCad PCB (etched and soldered by hand) and a worm-gear drive sized by Siemens NX structural FEA (~85 N simulated grip).

• 92% accuracy
• 414-644 ms latency
• ESP32 / C++ (PlatformIO)
• 4 grip patterns
• single sEMG channel
• hand-etched double-sided PCB
• FEA-sized drive (~85 N)

A wireless biosignal telemetry system: an ESP32 streams forearm EMG over Wi-Fi (MQTT) to a Raspberry Pi 5 running embedded Linux, which filters the signal with a 50 Hz notch, serves a live dashboard, forwards a decimated feed to AWS IoT Core over mutual TLS, and exposes the live muscle-activity level through a hand-written Linux kernel character driver (/dev/emgactivity). It is the networked-Linux and embedded-software companion to my STM32 gripper, applied to biosignals (the shape of remote biosignal monitoring).

• ESP32 / Wi-Fi + MQTT
• Raspberry Pi 5 (embedded Linux)
• Linux kernel driver (C)
• 50 Hz notch DSP
• AWS serverless (IoT Core, Lambda, DynamoDB)
• systemd + GitHub Actions CI