feasix.com

What It Is

Feasix AI focuses on a “foundation model for outdoor autonomous mobile robots,” aiming to use Physical AI to help mobile robots move beyond indoor environments. Its core thesis is that traditional SLAM mainly understands the world through points and lines in LiDAR point clouds, making it difficult to distinguish objects such as people, trees, and vehicles. Rule-based systems, meanwhile, rely on existing maps and fixed rules, and can easily fail when faced with construction, snow, ice, seasonal changes, or changes in lighting and weather.

Core Capabilities

According to the website, the RFM architecture takes RGB Camera and optional GPS as inputs, and uses a pre-trained robot Foundation Model, an SSL vision backbone, an imitation-learning Action Policy, and a contextual understanding module to output target waypoints. Its technical focus includes unlabeled self-supervised visual representation learning, learning driving behavior from expert Tele-op data, and generalization capabilities for unstructured outdoor environments. The company also emphasizes in-house hardware, remote operation capabilities, and a continuous data flywheel, all of which are critical to whether this type of robot model can keep improving over time.

Typical Use Cases and Business Model

Feasix AI is not simply selling software; it proposes a “human + robot” collaborative service model. In parking management, robots are expected to handle around 80% of the work and humans 20%; for cleaning and snow removal, the split is 60%/40%; for logistics and transport, 70%/30%; and for hazardous sites, 90%/10%. Covered scenarios include parking patrols and illegal vehicle detection in residential communities, cleaning and snow removal at factories, power plants, and campuses, logistics at airports and industrial parks, as well as hazardous object detection and disaster rescue.

Pricing and Integration

The website does not disclose pricing, free trials, procurement models, SaaS/API fees, or project-based delivery options. It also does not explain API, SDK, deployment environments, or how it integrates with customers’ existing robots or dispatch systems. Payment methods are not provided either, so commercial deployment requires direct contact for confirmation.

Pros and Cons

Its strengths are a clearly defined problem, a focus away from purely indoor robots and toward more difficult but higher-value outdoor scenarios, and a technically coherent roadmap combining self-supervised learning, imitation learning, and a closed loop of real operational data. The founding team also has experience in robot R&D, mass production, AI software, and parking systems. Limitations include the lack of quantified performance metrics, real customer case studies, safety certifications, failure scenarios, and explanations around data privacy and data ownership. The reliability of the so-called minimal RGB+GPS sensor setup also needs to be validated through real-world testing in rain, snow, nighttime conditions, strong backlighting, and mixed pedestrian-vehicle environments.

Who It’s For and Access from China

Feasix AI is better suited to B2B customers with needs in outdoor inspection, parking management, campus logistics, cleaning and snow removal, or hazardous-environment operations, as well as organizations seeking robotics technology partnerships. Access from mainland China, Chinese-language support, and local payment options are not disclosed and should be considered unknown. For domestic deployment in China, key factors to evaluate include network connectivity, after-sales response, on-site data compliance, and local alternatives.