Parkour, obstacle navigation, backflips...Today's robotic dogs are already at the peak of their athletic abilities.
But can a robot dog go beyond pre-programmed routines, understand natural language, and autonomously complete continuous tasks? This is the key to bringing embodied intelligence into the real world.
As a leading Physical AI company, Nullmax is extending AI capabilities validated in the real world into the broader physical world. It has fully self-developed a general-purpose Physical AI technology foundation — Nullmax Physical AI (NPI) — and uses it to build NPI-Brain,an “Embodied Brain”. It provides complete capabilities from perception, understanding, memory, and decision-making to motion and manipulation control, enabling machines to move from "seeing and understanding" to "autonomously interacting with the world."
In the latest real-world demonstration of “Smart Mobility Series”, MAX, a robot dog powered by NPI-Brain, transforms into an "all-around task performer." Powered by NPI Ultra, it demonstrates outstanding long-range task execution across indoor, building, campus, and street scenarios. From understanding natural language and retrieving memories to plan routes, to seamlessly navigating across floors and areas, MAX can listen, see, speak, think, and act - all without human intervention. Breaking through industry bottlenecks, MAX eliminates the need for pre-programmed task sequences, enabling autonomous reasoning and long-term memory, achieving an upgrade from "rule execution" to "autonomous cognition and continuous operation."
01 Natural Human-Robot Streaming Conversation — Full-Duplex Interaction, Instant Intent Understanding
NPI Ultra supports multimodal information interaction such as voice and text. A single sentence is all it takes to assign a task. The platform also supports multi-turn dialogue: when task requirements are unclear, the robot dog can proactively ask follow-up questions to clarify intent. Throughout the conversation, users can freely interrupt, interject, or speak while the robot listens, making communication as natural as interacting with another person.
02 Intelligent Mobility — Reinforcement Learning-Driven, Free Traversal Across All Scenarios
Intelligent mobility is one of the core capabilities of NPI Ultra. Driven by Nullmax's reinforcement learning algorithms at its core, it builds intelligent mobility that does not rely on the single paradigm of traditional SLAM. From indoor patrols to outdoor guiding, the robot dog can automatically recognize semantic locations such as meeting rooms, reception areas, and elevator lobbies, perform real-time dynamic obstacle avoidance and detouring, and intelligently traverse multiple terrains across all scenarios. When it deviates from its planned path, MAX can replan and proactively report back, keeping track of every step along the way.
03 Long-Term Memory — Persistent Online, Automatic Archiving
Traditional robots suffer from fragmented modules and a lack of "memory." Once the environment changes or the target is occluded, they struggle to complete closed-loop tasks.
NPI's tokenized scene memory capability unifies voice, vision, and status into the brain's context, building a closed-loop decision-making architecture with persistent memory and spatial cognition. This enables the robot dog to remember the environment, maintain task continuity, and continuously accumulate experience from historical execution.
04 Closed-Loop Execution — From "Single-Point Movement" to "Handling the Entire Journey"
Leveraging the high mobility of the quadruped body and the collaboration of NPI Ultra, the robot dog seamlessly integrates multiple skills, including semantic understanding, route planning, cross-floor navigation, visual person identification, and voice interaction. Transforming into an “all-around task performer” across office buildings, it independently completes the “last mile” and executes long-horizon tasks seamlessly from start to finish.
In the future, Nullmax will continue to push the boundaries of embodied intelligence with its industry-leading embodied brain, expand its capabilities across an ever-wider range of scenarios, and enable flexible adaptation to diverse robotic platforms. Ultimately, bringing to life the vision of “One Brain for All, Intelligence Transforming Everything.”