David Docherty, PhD

Robotics engineering for defense systems. System and test engineering as well as algorithmic design.

Developing a new model of emotion that allows robots to benefit from the increased survivability that emotions provided early animals. I am currently implementing on-line reinforcement learning in the form of classical conditioning, via MATLAB, to generate motor commands for a simulated robot but I hope to eventually develop the hardware as well. Outcomes: A simulated robot that can learn predictive emotional responses to stimuli. A novel mathematical model of the emotional/behavioural responses of simple animals. Creation of a simulated mobile sensor platform capable of reactive and predictive navigation.

PhD student at Maersk McKinney Moller Institute: Biorobotics, paired with Biology department: Bioacoustic group. Research and exploration of search algorithms to ascertain the best method that a drone or drones can use to find a bat hunting in a large area, as well as the development of audio capture hardware and drones. This included the creation of detailed mathematical models of the flight behaviour of both drones and bats as well as the movement of the bat's audio emission through space. Using these flight models, I developed a series of optimised search algorithms using MATLAB and generated flight paths that could be uploaded to an onboard computer on the UAV. This onboard computer was a modified audio processing board that was designed to grab ultrasound signals from a large microphone array at an extremely high framerate - my developments allowed the audio board to communicate with UAV's and update flight commands based on the audio signals that were being processed in parallel. For efficient testing I also developed a bat analog that consisted of a variably rotating ultrasonic loudspeaker. Compressing the figure-8 style flight path of the bat's audio emission into a single point rotating at different speeds and directions thus, creating a portable bat emitter that's received signal was significantly similar to a real bat. Outcomes: Efficient search algorithms taking into account physical limitations such as battery life and flight speed. Transmission of mathematical search algorithm to drone onboard computer to drone commands. Mathematical models of bat flight. A bat analog device. A robust, custom, audio processing board. 3 semesters of teaching experience.

Developer of proactive human-robot interaction using audio processing, speech recognition, sensor fusion and motor control. In this position I implemented principles such as the Lizard Ear Model and Beamforming to accurately predict the location of a speech source, as well as speech recognition software, to orient a health-care robot towards a potential interlocutor. The audio was paired with computer vision data using sensor fusion through on-line learning to generate motor commands that would guide the robot's motion. The distributed system of audio processors, Nvidia Jetsons, and the robot platform were organised using the Robot Operating System (ROS). Outcomes: Robust speech localisation software and hardware using C++ and MATLAB. Sensor fusion of audio and vision, using ICO/ISO learning, to produce motor commands. Automatic orientation of attention towards an interlocutor.

Developer on The Robot Assisted Laser Micro-Surgery project. My work included the development of precision control systems for a surgical laser, by fusing sensor feedback and user input. My implementation of a graphical tablet allowed surgeons to draw a cutting path directly onto a magnified live-feed of a patient’s larynx which could then be discussed, edited, and finalised - the surgeon could then activate the laser which would then cut along the path that was in reality only millimeters in size. Outcomes: A motor control board, safety system, and tablet interface for a surgical laser. Micron level precision of guidance motors. More accurate, safer, more collaborative larynx surgeries. Competencies: Medically/publically safe circuit design. Medical level software and hardware documentation. Microprocessor and micro controller programming and interfacing.

Explored and developed search algorithms for drones to find bats hunting in the wild. The aim of the research was to be able to detect and record the ultrasound of wild bats whilst they are engaging in previously unobserved behaviours. The rapid testing of many different algorithms, under different circumstances such as different flight speed and battery life, was achieved using a mathematical model of both the bat and drone in Matlab. Many real-world tests were attempted but were ultimately not undertaken due to the error sensitivity of drones (small errors leading to substantial delays) but bat-recording drones were successfully implemented manually.

Final thesis on the development of complex emotions, from real world stimuli, by a robot using recursive neural nets. Masters program included: Embedded AI, VHDL and FPGAs, programming, hardware design, automation, human robot interaction, and sensor theory.

BSc thesis and project developing a modular centipede robot of undetermined length, whereby a series of body segments (including 2 legs) could be attached end-on-end to create a walking robot with any number of leg pairs. The robot was designed to behave in an animalistic manner when navigating and being interacted with. Bachelors program included: Programming in C, C++, Python, Assembly, and more; VHDL and FPGAs; hardware/circuit design; and sensor theory.