Polymath Robotics
"Autonomy & Safety Systems for Off-Highway Vehicles"
San Francisco autonomy software startup building modular full-stack safety and navigation systems for off-highway vehicles via REST API, backed by Y Combinator.
Polymath Robotics is a San Francisco-based autonomy software company founded in September 2021 by Stefan Seltz-Axmacher and Ilia Baranov, and publicly launched in July 2022. The company is a Y Combinator Summer 2022 cohort member. CEO Stefan Seltz-Axmacher previously co-founded Starsky Robotics, which became the first company to drive an unmanned truck on a public highway; co-founder Ilia Baranov brings prior experience from Clearpath Robotics and Amazon Lab 126.
Polymath Robotics builds core autonomy and safety systems for off-highway vehicles, designed to be retrofitted onto existing equipment or integrated directly by original equipment manufacturers (OEMs). The company's platform enables full-stack autonomous operation across a wide range of heavy industrial machinery including agricultural vehicles, mining trucks, construction equipment, logging machines, and marine vessels. The company serves OEMs, fleet operators, and defense organizations.
The primary product is the Autonomous Navigation Modules platform — a toolkit of over 40 distinct modular autonomy components that can be deployed individually or combined into comprehensive navigation systems. Key modules include GPS-denied trailer localization (intelligently estimating position and orientation without satellite signals), LiDAR fusion (transforming raw, noisy multi-beam LiDAR data into clean, ready-to-use point clouds compatible with any LiDAR type), mission planning, perception layers, and integrated safety systems. The platform bundles AI, machine learning, controls engineering, ROS (Robot Operating System), and industry-standard safety best practices.
Vehicles equipped with Polymath's systems are commandable via a REST API, enabling customers to issue navigation commands and monitor vehicle state using any preferred programming language. The company also provides robotics DevOps tooling including CI/CD pipelines, testing frameworks, and monitoring infrastructure for autonomous system development teams. Polymath uses LiveKit for real-time remote operation of heavy machinery over network connections.
The platform addresses industries with hazardous, repetitive, or labor-scarce operational tasks including agriculture, mining, logistics, construction, and defense — providing the fastest and most cost-effective path to vehicle automation without requiring operators to build foundational autonomy technology from scratch.
Primary type & automation activities this supplier delivers:
Autonomous Navigation Modules
- Launch Year: 2022
Toolkit of 40+ modular autonomy components for off-highway vehicles, commandable via REST API with GPS-denied navigation.
Contact Polymath Robotics
WEBSITE
https://www.polymathrobotics.comPHONE
628-600-2680HEADQUARTERS
United States
Company Facts
Founded
2021
Primary Role
Software/Algorithm
Company Size
-
Primary Region
North America
Annual Sales
-
Funding Stage
-
Funding Total
-
Keywords
- Polymath Robotics
- off-highway
- autonomous-navigation
- REST-API
- ROS
- modular-autonomy
- heavy-machinery
- Autonomous Vehicle / UGV
- Software/Algorithm
- Autonomous Vehicle/UGV
- Control & Software
- Navigation & Positioning
- Agriculture
- Manufacturing
- Logistics & Supply Chain
- Defense & Security
- Special / Extreme Environments
- Mining
Camtastic GigE Cameras Deliver Reliability and Precision in Sugar-Cube Format
Camtastic has announced its GigE cameras, which provide exceptional image processing capabilities through a GigE-Vision platform. This platform ensures seamless integration and rapid deployment within a comprehensive GigE ecosystem. The introduction of these cameras is significant as they enhance the reliability and precision of image processing tasks, catering to various industrial applications. The GigE-Vision platform allows for efficient connectivity and interoperability, which is crucial for modern automation and imaging systems. Looking ahead, industry professionals should monitor the adoption of Camtastic's GigE cameras in various sectors that require high-performance imaging solutions. No further timeline was disclosed at the time of publication.
Allgemein
Manufacturers Face Challenges in Implementing AI on the Shop Floor
A recent report highlights that while 75% of manufacturers have integrated AI into their enterprise strategies, only 20% of these initiatives achieve their business goals. This gap underscores the need for better integration of AI with operational systems and human decision-making processes. The Infosys Manufacturing Tech Index: AI Pulse indicates that the success of AI in manufacturing is closely tied to the convergence of information technology (IT) and operational technology (OT). Effective AI applications require a cohesive operating view that combines data from various sources, including production telemetry, enterprise systems, and maintenance records. Looking ahead, manufacturers must focus on bridging the IT/OT divide to enhance AI's impact on production decisions. The integration of predictive maintenance, digital twins, and computer vision systems will be crucial for enabling real-time responses and improving operational efficiency. No further timeline was disclosed at the time of publication.
Nucleus Demonstrates Humanoid Robot Performing Factory Tasks with 60% Autonomy
Nucleus has released nearly two hours of uncut footage showcasing its humanoid robot performing factory tasks, revealing a reliance on human assistance. The demonstration, shared by CEO Melvin Schwarz on October 1, indicated that the robot operated approximately 60% autonomously and 40% through teleoperation, although the method for calculating these percentages was not disclosed. This demonstration is significant as it highlights the ongoing challenge of maintaining useful robotic work over extended periods. Nucleus aims to prove that its robots can effectively integrate into industrial workflows, running shifts of four to six hours while completing tasks from start to finish, albeit with necessary human intervention. Looking ahead, Nucleus is focused on refining its AI systems through real-world applications, capturing data from both successful operations and instances requiring human guidance. This feedback loop is essential for improving the robot's capabilities, particularly in adapting to varying factory conditions and ensuring consistent performance across different tasks.
AI and Robotics
Podcast Discusses Smarter Predictive Asset Management and Its Operational Benefits
In a recent episode of Automation World Gets Your Questions Answered, Chris McNamara interviews Steve Kaminski from Banner Engineering. Kaminski discusses how condition monitoring transforms machine data into actionable insights, enabling manufacturers to maintain efficiency and prevent failures. He emphasizes starting with critical equipment like compressors and pumps for optimal results. This conversation is significant as it underscores the importance of condition monitoring in enhancing operational efficiency. By utilizing Banner’s Snap Signal ecosystem, manufacturers can integrate various sensors into a cohesive data network, which aids in early detection of mechanical issues and optimizes maintenance schedules. This proactive approach can lead to reduced downtime and lower operational costs. Looking forward, Kaminski anticipates advancements in user-friendly systems that will facilitate both basic monitoring and automated maintenance workflows. No further timeline was disclosed at the time of publication.
Sponsored
FMC³ Robotics, Vector Informatik, and PEM Motion Form Joint Venture for Embodied AI Development
At the Humanoid Robots Summit 2026 in Stuttgart, FMC³ Robotics, Vector Informatik, and PEM Motion signed a memorandum of understanding to develop a standardized data infrastructure for industrial embodied AI. This collaboration aims to enhance the collection, processing, training, validation, deployment, and continuous improvement of AI-powered robotic systems in manufacturing environments. The partnership is significant as it combines the strengths of the three companies in production engineering, standardized industrial data technologies, and embodied AI. As robots transition from pilot projects to real production, the need for high-quality data and a robust infrastructure becomes critical for reliable performance in manufacturing settings. Looking ahead, the focus will be on creating a unified data infrastructure based on standardized protocols to support multimodal data collection and processing. No further timeline was disclosed at the time of publication.
Understanding the Concept of 'Open' in Industrial Automation
The term 'Open' has become ubiquitous in industrial automation, but its meaning is often vague. It encompasses various measures, and suppliers may be open in some aspects while closed in others. Buyers should ask specific questions to discern the true openness of a platform or system. Historically, Open Automation 1.0 focused on interoperability, allowing different systems to communicate without custom drivers. With standards like OPC UA and MQTT, this challenge has largely been addressed. Open Automation 2.0 shifts the focus to application freedom, enabling users to run third-party software on devices, which enhances flexibility and reduces costs. As suppliers navigate the balance between openness and security, buyers must recognize that openness involves trade-offs. While opening systems can increase vulnerability, established standards like ISA/IEC 62443 provide frameworks for maintaining security. Understanding these dynamics is crucial for making informed decisions in industrial automation.
Exploring the Impact of Manipulated Inputs on Robot Safety Functions
VicOne LAB R7 investigates how manipulated inputs can affect robot behavior and safety. Mobile robots, collaborative arms, and humanoids rely on accurate environmental data to function safely. When inputs are incorrect, whether due to faults or deliberate manipulation, robots may respond inappropriately, raising concerns about safety protocols. Understanding the implications of these findings is crucial for safety and security teams. Research indicates that adversarial inputs can lead to unintended robot actions, as demonstrated in various tests involving vision-language-action models and audio manipulation. These studies highlight the need for robust safety assessments that account for both accidental faults and deliberate cyber threats. Looking ahead, organizations must evaluate their safety measures to ensure they can detect manipulated inputs effectively. The potential for attackers to exploit vulnerabilities in robot systems necessitates a reevaluation of risk assessments and safety protocols. No further timeline was disclosed at the time of publication.
Sponsored Content VicOne
Dürr Introduces Four New EcoRS Generation 2 Robots for Automotive Paint Shops
Dürr has unveiled four new industrial robots tailored for automotive paint shops, with payload capacities ranging from 20 kg to 150 kg. The EcoRS Generation 2 series is designed for various applications, including sealing, cleaning, and material handling, and features a reach of up to 3,100 mm. This standardized platform aims to streamline paint-shop automation planning and maintenance. The introduction of the EcoRS Generation 2 robots is significant as it establishes a standardized solution for sealing applications, enhancing process quality and availability while reducing complexity throughout the equipment's lifecycle. Martin Wurst, product manager at Dürr, emphasized the benefits of this new platform for customers, particularly in achieving consistent seams across vehicle bodies. Looking ahead, Dürr's focus on integrating robot mechanics, control systems, and application technology is expected to improve path accuracy and operational flexibility. The redesigned EcoRS platform, with its internal component relocation, not only minimizes interference but also simplifies cleaning and reduces space requirements. No further timeline was disclosed at the time of publication.
Physical AI Requires a Fourth Verb: Verification for Production Accuracy
Physical AI is commonly defined as machines that perceive, reason, and act in the physical world. However, this definition lacks a crucial fourth verb: verify. In production environments, acting does not equate to producing conforming parts, as demonstrated by a humanoid robot at BMW's Spartanburg plant, which successfully placed over 90,000 components with a 99% success rate per shift. The importance of verification lies in ensuring that produced parts meet specified tolerances. While the robot's execution rate is impressive, it does not guarantee that the assemblies are within acceptable limits due to various factors such as fixture wear and thermal drift. The introduction of in-line dimensional inspection has made it feasible to measure critical characteristics and compare conformance rates against reported success rates, highlighting the need for this fourth verb in the definition of Physical AI. Going forward, the industry must focus on integrating verification into the acceptance criteria for Physical AI systems. This shift will enhance the reliability of production processes, ensuring that machines not only report their actions but also the quality of their outputs. No further timeline was disclosed at the time of publication.
Factory / Digital Transformation
The Challenge of Implementing Spatial and Agentic AI in Smart Factories
Smart factories are equipped with advanced technologies like artificial intelligence, spatial computing, and autonomous robotics, yet many manufacturers are not fully leveraging these capabilities. A 2025 State of AI in Manufacturing Survey reveals that while over 77% of manufacturers have adopted AI, 56% are uncertain about their systems' readiness for complete integration. This indicates that the gap between technology adoption and operational value stems from implementation challenges rather than technological limitations. Spatial AI enables machines to understand and act within three-dimensional environments, while agentic AI allows systems to autonomously pursue goals across workflows. This combination could transform factory operations into dynamic, self-correcting systems. However, the effectiveness of these technologies is hindered by fragmented data ecosystems, where inconsistent and poorly contextualized data leads to misleading AI outputs. The need for a unified data layer is critical for effective smart factory operations. Looking ahead, manufacturers must prioritize the integration of legacy systems and establish shared data standards to fully realize the potential of smart factories. A Deloitte survey indicates that 41% of manufacturing executives plan to invest in automation hardware in the next two years, but these investments will only yield benefits if the underlying data infrastructure is robust enough to support them. No further timeline was disclosed at the time of publication.
Factory / Analytics
WRC 2026 Robot Systems: 149 Companies That Build Robots
149 of the 248 classified WRC 2026 exhibitors build finished robots. 93 build humanoid or legged machines. The audited humanoid sales number is one unit.
ROBOTTODAY Weekly July 20 – 24, 2026
Your weekly robotics briefing: AGIBOT unveils four new robots at WAIC 2026, the NDAA bars Chinese humanoids from the U.S. military, SoftBank eyes Gravis Robotics for its Roze vehicle, and Anduril reveals its Thunder autonomous attack tiltrotor at Farnborough.
US Restrictions on Chinese Humanoid Robots: The Full Map
Every US restriction on Chinese humanoid robots — NDAA Sec. 163, GUARD Act, the 1260H list — plus the supply-chain math and a 4-tier map of decoupling.
FUTURE WARFARE | The Geopolitical Competitive Landscape
Military robotics is no longer a U.S.-centric story. Global defence spending crossed $2.75 trillion in 2026; defence-tech VC hit a record $49.1 billion in 2025; and a multipolar robotics order is crystallising in real time — with China’s 15th Five-Year Plan institutionalising civil-military fusion, Europe’s €381 billion defence budget financing its own AI champions, and Turkey and Israel reshaping the export market with affordable, combat-proven systems. Part Five of the Future Warfare Series maps the country-by-country competitive landscape, the collapsing regulation race, and the investment forces reshaping who builds — and who controls — the autonomous weapons of the next decade.
FUTURE WARFARE | Cross-Domain Integration - Integration How Land, Sea, Air and Software Are Converging Into a Single Kill Chain
Cross-domain robotics is reshaping modern warfare as drones, UGVs, AI targeting systems, and autonomous logistics converge into a single networked kill chain across land, sea, air, and software.
GLOBAL ROBOTICS INDUSTRY Comprehensive Sector Overview
A comprehensive global robotics industry overview for 2025, analyzing 14 key market segments — from industrial and humanoid robots to logistics AMRs, military systems, and service robots — with expert market sizing, US case studies, five-year forecasts, and bilingual English-Chinese analysis for executives and investors.
FUTURE WARFARE | Land - Unmanned Ground Vehicles — From Logistics Mules to Front-Line Robots
Ukraine fielded 15,000 unmanned ground vehicles in 2025 — surpassing its own procurement targets by over 100 percent. A single UGV held a front-line position for 45 days. Up to 90 percent of supplies to Pokrovsk now move by robot, not truck. The U.S. Army cancelled its $3 million Robotic Combat Vehicle and is starting over. China is deploying $3,000 robot dogs in PLA urban-warfare exercises. This article maps the emerging UGV battlefield — from Ukraine’s garage-built logistics fleet to the Pentagon’s stalled combat-vehicle ambitions, and from Ghost Robotics’ Vision 60 to the quadruped proliferation problem no treaty has yet addressed.
EU Bets €1.5B on Drones and AI: Inside EDIP's 2026–2027 Defence Industry Programme
The EU's €1.47 billion European Defence Industry Programme (EDIP) 2026–2027 targets drones, counter-UxS systems, AI-enabled autonomy, and loitering munitions across six major funding calls — allocating over €232 million directly to unmanned systems and related electronics supply chains.
FUTURE WARFARE | The Autonomy Spectrum
How modern weapons shattered the concept of human control — from kill-switch confirmation to AI systems that select targets independently.
EKA Mobility Secures $57 Million from NIIF’s India-Japan Fund to Drive Leadership in Electric Commercial Vehicles
EKA Mobility raises $57 M from the India-Japan Fund (NIIF) to accelerate electric commercial vehicle growth and electrification leadership in key markets.
