GPU Computing for Cognitive Robotics
The technical foundation: iCub humanoid learning, computer vision, neural networks, and GPU acceleration.
Plymouth PhD thesis recordSources
The original papers, institutional records, talks and preserved material behind my work. Start with four primary sources, or jump straight to a subject.
For full academic citations, DOI records, available PDFs and BibTeX, open the research bibliography.
Start here
If you only open four things, open the thesis, TEDx, the ESA rover report, and the CUDA/NVIDIA sources.
The technical foundation: iCub humanoid learning, computer vision, neural networks, and GPU acceleration.
Plymouth PhD thesis recordA public telling of the early story: robots, restart, study, and the long route into artificial intelligence.
TEDxBratislava talk pageThe public ESA record for rover simulation, active vision, and island-model neuro-controller work with the Plymouth group.
Read the preserved ESA report Read the project team’s accountThe public CUDA Spotlight record connecting iCub research, GPU acceleration, and the early GPU robotics period.
NVIDIA CUDA SpotlightFamily testimony
The original photographed typescript, three complete illustrated editions, and the documentary made from this family inheritance. Pavol’s written account and authentic family photographs are distinguished from the film’s AI reconstructions.
Further evidence
Press and institutional mentions
Selected contemporary coverage and later mentions. These records preserve how the work was described at the time; the case studies and bibliography give the technical context.
Staromestské noviny and SITA named me in their previews of the 5 July event, describing the cognitive-robotics work behind my talk.
The original Slovak programme includes my speaker biography on PDF page 24 and the event timetable on page 5.
Read the official programmePlymouth’s article about its research and innovation strategy identifies me in a photograph playing ball with a humanoid robot at CRNS.
Read the university articleIn an interview with Riccardo Oldani, my supervisor Angelo Cangelosi recalled that robotics researchers shared my video comparing GPU and CPU performance.
Read the interview in ItalianRiccardo Oldani’s article discusses my route into robotics and GPU computing. It is a historical media account; the thesis and research papers document the scope of the work.
Read the Italian coverageProfiles and indexes
Bibliographic indexes and my public profiles. Older third-party biographies may describe roles held at the time; the CV gives the reconciled career chronology.
Foundations
These are the source materials behind the iCub, Aquila, CUDA, and rover-simulation work.
University of Plymouth PhD thesis on cognitive robotics, iCub humanoid learning, neural networks, action-language acquisition, computer vision, and GPU acceleration. The repository also preserves four supplementary experiment videos.
Aquila 2.0 framed the robotics work as reusable infrastructure: heterogeneous CPU/GPU modules, loosely coupled GUIs, and scalable tools for cognitive-robotics experiments.
Aquila 2.0 publication recordPeer-reviewed IJCNN record for the Aquila cognitive and neuro-robotics toolkit, connecting open software with GPU-accelerated robotics experiments.
Aquila open-source toolkit recordGTC 2012 poster from the pre-CNN-boom GPU-compute era: iCub action acquisition with multiple time-scales recurrent neural networks, Aquila, CUDA, and measured GPU speedups.
NVIDIA GTC 2012 poster PDFNVIDIA-hosted GTC Express presentation of Aquila as a GPU-accelerated toolkit for cognitive and neuro-robotics research.
Open NVIDIA slidesPhotograph of Martin Peniak presenting the GPU-accelerated cognitive-robotics and Aquila work at NVIDIA GTC in 2014.
Robotika.SK/STU Bratislava announcement for a talk on iCub action-language learning, MTRNN models, Aquila, iTALK, and CUDA acceleration.
Robotika.SK 2011 talk pageLecture slides by Martin Peniak and Anthony Morse, presented in Plymouth on 11 June 2010. Historical teaching material on the robot, its software and simulation tools.
Open the teaching slidesOpen-source project record for Aquila: software architecture for cognitive robotics, with preserved releases and source history for the modular CPU/GPU toolkit.
The AISB conference report records the 2011 best paper award for Aquila: Massively Parallelised Developmental Robotics Framework.
AISB Quarterly conference reportI co-authored the ESA study report. Plymouth lists Angelo Cangelosi, Davide Marocco and me as project leads and credits Barry Bentley for his assistance.
My GPUGrid account dates from 9 March 2010 and lists the University of Plymouth team. I contributed computing time for scientific simulations.
GPUGrid contributor recordSourceForge preserves the simulator used for terrain, sensing, active-vision, and neuro-controller experiments around the Plymouth and ESA work.
Photograph from the ESA-linked Mars rover collaboration, showing Martin Peniak with the rover research platform in the test setting.
Group photograph from the ESA-linked Mars rover collaboration around Evolution in Robotic Islands, rover simulation, active vision, and neuro-controller research.
Photograph with colleagues at ESA ACT in Noordwijk during the 2010 Mars rover collaboration, with the simulator work visible in the room.
ESA-hosted final report for the Evolution in Robotic Islands study, with Cangelosi, Marocco, Peniak, and ESA collaborators.
ESA Robotic Islands final report and preservation recordPublic article on the Plymouth and ESA collaboration, Mars Rover Simulator, active vision, and island-model rover-controller work.
ERCIM article on autonomous Mars roversConference paper on autonomous navigation, rover simulation, and neuro-controller optimization for planetary exploration.
Author manuscript (PDF)Workshop paper connecting evolutionary robotics, camera control, and planetary-rover navigation under unknown terrain constraints.
Workshop paper, author copy (PDF)The preliminary Mars rover model presented at the 10th ESA ASTRA Workshop in Noordwijk, 11-13 November 2008, before the formal ACT study.
ASTRA author manuscript (PDF)Frontiers conference abstract for the MTRNN action-acquisition work. It is preserved here as an abstract/poster record rather than labelled as a full peer-reviewed paper.
Frontiers MTRNN action-acquisition recordResearch on making larger action-learning experiments practical through GPU acceleration.
Manchester GPU action-learning recordAutonomous Robots 43, 1271-1290 (2019), first published online in 2018. The article connects embodied input, learning, and noun-verb generalisation.
Journal article and DOIPeer-reviewed Topics in Cognitive Science article, 6(3), 534-544 (2014), on individual, social, and linguistic learning in developmental robotics.
At Plymouth I continued my work on robot action and language in POETICON++. The university lists Angelo Cangelosi, Tony Morse and me on the Plymouth team.
Planetary-rover simulation and controller-design work related to the ESA rover-control research.
ESA study reportUniversity publication record for simulated Mars rover explorer work and controller/sensing co-evolution.
Manchester rover co-evolution recordArchive PDFs
Metadata-stripped archive derivatives for older papers, posters, and talk files that are part of the public research record. External canonical links remain preferred where available.
Early ESA workshop source for rover simulation and autonomous controller design.
Local archive PDFRover simulation paper on controller and sensing co-evolution.
Local archive PDFPoster source for active vision, Mars rover simulation, and ESA-linked research.
Local archive PDFWorkshop paper on camera control and rover navigation under unknown conditions.
Local archive PDFPaper on evolving neuro-controllers for planetary rover control.
Local archive PDFThe TAROS 2010 paper on evolving rover controllers and sensing abilities.
Local archive PDFOne-page source tied to the PCCAT best paper award certificate.
Local archive PDFHumanoid robotics source around action, language, and embodied learning.
Local archive PDFArchive copy of the Aquila open-source GPU-accelerated cognitive robotics paper.
Local archive PDFSurviving author manuscript; publication venue and date remain unconfirmed. See the bibliography note.
Local archive PDFNVIDIA GTC poster source for early GPU-accelerated action acquisition.
Local archive PDFArchive slide deck for the NVIDIA GTC Express webinar.
Local archive PDFArchive source for the Aquila 2.0 cognitive-robotics software architecture paper.
Local archive PDFStatic copy of the Robotika.SK talk announcement for action and language in humanoid robots.
Local archive PDFSlide archive for active vision, rover simulation, and camera movement as part of perception.
Local archive PDFProfiles, articles, and talks
External profiles, interviews, articles, technical sources, and talks sit beside the research record.
NVIDIA/CUDA spotlight around GPU acceleration and cognitive robotics research.
NVIDIA CUDA SpotlightTEDx Bratislava talk connecting the personal story, artificial intelligence, humanoid robotics, and engineering optimism.
TEDxBratislava speaker pageEdited transcript of a live Second Home talk connecting the Slovak builder-to-robotics story with GPU computing, iCub, NVIDIA, ESA, and visual search.
Second Home talk transcriptUniversity of Plymouth profile from Connect, Winter 2012, issue 4, pages 18-19, preserving the route from arrival in Plymouth to the BSc, ESA, iCub, NVIDIA, and POETICON++.
Open preserved magazine scansArchived first-person NVIDIA blog article on Aquila, CUDA, iCub, and the early GPU-computing direction of the PhD.
Open Internet Archive copyOkinawa Institute of Science and Technology preserves and attributes the iCub/MTRNN demonstration to Martin Peniak at the University of Plymouth.
Open OIST recordUniversity profile focused on the PhD topic, ESA rover collaboration, NVIDIA internship, Aquila, TEDx, and the early NVIDIA CUDA Teaching Center contribution at Plymouth.
The original university page is no longer available. This dated Internet Archive copy preserves the historical profile; its employment details describe that period.
University of Plymouth profile (archived 15 April 2026)Photograph with Martin Peniak, supervisor Angelo Cangelosi, colleagues, student and co-author Barry Bentley, and the iCub platform in the Plymouth robotics lab.
Plymouth iCub research group photographOfficial RobotCub summer school page for VVV ’09, paired with a photograph of Martin Peniak with iCub during the 2009 summer school.
Photograph of the Science Museum London iCub display showing Martin Peniak’s robotics video inside the exhibit.
Science Museum iCub exhibit photographPhotograph of Martin Peniak and the iCub humanoid robot in a Manchester space-exhibition setting.
Manchester space-exhibit iCub photographCollage from a free telescope evening at Plymouth Hoe, connecting the Plymouth work-study years with astronomy, astrophotography, and public curiosity.
Plymouth Hoe telescope night photographClipping from Slovak press showing the iCub humanoid robotics work alongside the rover and space-research work.
Slovak press clippingBibliographic index for Martin Peniak publications and research records.
DBLP publication profileResearch profile listing publications and citation context.
ResearchGate profileAcademic search profile for research discovery and citation links.
Google Scholar citation profileArchive
Some older blog and project links now resolve to concise internal pages so useful references do not disappear.
GPU-accelerated cognitive robotics toolkit context for iCub, CUDA, and modular CPU-GPU experimentation.
Aquila archive pageOpen-source rover simulation and ESA planetary robotics research around active vision and island-model optimization.
Mars Rover Simulator archive pagePreserved technical note for old developer citations to global CUDA thread indexing examples.
CUDA indexing noteStable context for Up Squared, Intel Myriad X, Google TPU, and early camera-as-computer prototypes now lives on the AI cameras page.
AI cameras pageAI cameras / applied vision sources
These source links connect the robotics/GPU work to later camera-as-computer systems, edge inference, Generator, and spatial calibration.
Cortexica material around ingress workflows and edge inference using Intel Movidius/Myriad-style hardware.
Open Cortexica edge-vision materialPublic Centre for Earth Observation Instrumentation record for the OVERPaSS project, naming Cortexica among the partners exploring GPU-enabled image analysis for future optical imaging satellites.
CEOI OVERPaSS project recordPublic NVIDIA Jetson Developer Challenge entry for a horizontally scalable edge-compute system built around Jetson TX2, CUDA kernels, and deep-learning inference.
NVIDIA challenge project recordMartin joined AnyVision as Senior Innovation Engineer in October 2019. His innovation work continued through the AnyVision-to-Oosto rebrand and Oosto’s acquisition by Metropolis, where he remains Senior Innovation Engineer.
Public arXiv record from the Cortexica-era AI cameras / applied vision period, connecting scalable image analysis, sandboxed execution, and practical vision systems.
arXiv large-scale image-analysis paperGuest article on UP Squared, Movidius AI Core, edge inference, and the bridge from GPU robotics to AI camera prototypes.
UP Board edge-computing articleVideo overview tying applied computer vision, edge AI, and working camera systems into one industry-era arc.
The Path of Innovation videoVideo showing early AI camera and edge-perception prototypes.
CortexiCAM Myriad-X videoVideo of synthetic worlds, known cameras, controllable truth, and repeatable perception experiments.
Generator v1.0 synthetic-world videoThe site page connects edge vision, synthetic truth, and the multi-view fused-cloud precursor.
AI cameras page