📌 TOPINDIATOURS Update ai: Watch: China’s new hybrid robot dog sprints up stairs a

A new video released by Pudu Robotics is drawing attention to how far industrial quadruped robots have come in combining speed with mobility. The roughly 20-second clip shows the company’s PUDU D5 robot approaching a multi-section staircase, transitioning from wheels to legs, and climbing three separate stair flights with short flat stretches in between.

According to the company’s YouTube caption, the footage is shown in real time, with no edits or speed manipulation, and highlights a claimed climbing speed of 1.5 metres per second (nearly 5 feet per second).

The sequence is simple. The robot wheels forward on flat ground, raises its legs in a motion similar to a dog climbing stairs, and then resumes rolling where the terrain allows.

What stands out is not just the ability to climb stairs, something already demonstrated by other quadruped robots, but the speed and fluidity with which the D5 switches between locomotion modes.

The video positions this hybrid wheel-leg approach as a practical solution for environments that combine smooth surfaces with sudden elevation changes.

A hybrid approach to real-world mobility

The video comes one month after the unveiling of the PUDU D5 Series, when the Shenzhen-based company introduced two configurations. A fully legged version and a wheeled variant optimized for mixed terrain. In its press release announcing the series, Pudu framed the robots as tools for outdoor and industrial environments that remain difficult to automate, including areas with uneven ground, staircases, slopes, and large, multi-level layouts.

According to the company, the D5 series is designed for autonomous operation in such spaces, with onboard computing based on a dual-processor setup using NVIDIA’s Orin platform alongside an RK3588 chip, delivering up to 275 TOPS of computing performance.

That processing power is intended to support real-time simultaneous localization and mapping, 3D reconstruction, obstacle avoidance, and path planning, allowing the robot to operate without constant human supervision.

Pudu says the robot relies on a 360-degree perception system combining four 120-degree fisheye cameras and dual 192-line LiDAR sensors, generating dense 3D point clouds for navigation.

In controlled demonstrations, the company claims the D5 can map and navigate areas as large as one million square meters and travel up to 14 kilometers on a single charge, figures aimed at large facilities such as airports, metro systems, and industrial campuses.

Positioning in a crowded robot-dog field

Quadruped robots capable of climbing stairs are not new. Boston Dynamics’ Spot, one of the most recognizable robot dogs, has long demonstrated careful stair-climbing using its legs alone, prioritizing stability and balance over speed.

Chinese competitors such as Unitree Robotics and Deep Robotics have also pushed quadrupeds into more practical settings, though public demonstrations of wheeled-leg hybrids tackling stairs at speed remain limited.

Pudu’s approach appears to emphasize efficiency in mixed environments, using wheels wherever possible and reserving legged motion for obstacles such as stairs or steps up to 25 centimeters high.

The company claims the D5 can reach speeds of up to 5 metres per second on flat surfaces and climb slopes of up to 30 degrees, while carrying payloads of up to 30 kilograms.

Speaking during the product launch in December, founder and chief executive Felix Zhang described the D5 as part of a broader strategy. “The PUDU D5 series represents a significant leap forward in our vision for robotics,” Zhang said. “For Pudu, the D5 is not merely a new product, but a milestone in our mission to push robots deeper into more applications.”

From demonstrations to deployment

At a Tokyo robotics exhibition last month, the D5 drew crowds as it rolled through halls and negotiated steps, highlighting Pudu’s push beyond indoor service robots into heavier-duty roles. The company has said it has sold more than 100,000 robots globally across its product lines and is preparing for wider international expansion.

Whether the D5’s fast, hybrid stair-climbing means reliable performance outside controlled demonstrations remains to be seen. The new video is a clear signal of where Pudu believes industrial quadruped robots are headed. Faster, more adaptive, and designed for the messy transitions in real-world environments.

🔗 Sumber: interestingengineering.com

📌 TOPINDIATOURS Eksklusif ai: ByteDance Introduces Astra: A Dual-Model Architectur

The increasing integration of robots across various sectors, from industrial manufacturing to daily life, highlights a growing need for advanced navigation systems. However, contemporary robot navigation systems face significant challenges in diverse and complex indoor environments, exposing the limitations of traditional approaches. Addressing the fundamental questions of “Where am I?”, “Where am I going?”, and “How do I get there?”, ByteDance has developed Astra, an innovative dual-model architecture designed to overcome these traditional navigation bottlenecks and enable general-purpose mobile robots.

Traditional navigation systems typically consist of multiple, smaller, and often rule-based modules to handle the core challenges of target localization, self-localization, and path planning. Target localization involves understanding natural language or image cues to pinpoint a destination on a map. Self-localization requires a robot to determine its precise position within a map, especially challenging in repetitive environments like warehouses where traditional methods often rely on artificial landmarks (e.g., QR codes). Path planning further divides into global planning for rough route generation and local planning for real-time obstacle avoidance and reaching intermediate waypoints.

While foundation models have shown promise in integrating smaller models to tackle broader tasks, the optimal number of models and their effective integration for comprehensive navigation remained an open question.

ByteDance’s Astra, detailed in their paper “Astra: Toward General-Purpose Mobile Robots via Hierarchical Multimodal Learning” (website: https://astra-mobility.github.io/), addresses these limitations. Following the System 1/System 2 paradigm, Astra features two primary sub-models: Astra-Global and Astra-Local. Astra-Global handles low-frequency tasks like target and self-localization, while Astra-Local manages high-frequency tasks such as local path planning and odometry estimation. This architecture promises to revolutionize how robots navigate complex indoor spaces.

Astra-Global: The Intelligent Brain for Global Localization

Astra-Global serves as the intelligent core of the Astra architecture, responsible for critical low-frequency tasks: self-localization and target localization. It functions as a Multimodal Large Language Model (MLLM), adept at processing both visual and linguistic inputs to achieve precise global positioning within a map. Its strength lies in utilizing a hybrid topological-semantic graph as contextual input, allowing the model to accurately locate positions based on query images or text prompts.

The construction of this robust localization system begins with offline mapping. The research team developed an offline method to build a hybrid topological-semantic graph G=(V,E,L):

• V (Nodes): Keyframes, obtained by temporal downsampling of input video and SfM-estimated 6-Degrees-of-Freedom (DoF) camera poses, act as nodes encoding camera poses and landmark references.
• E (Edges): Undirected edges establish connectivity based on relative node poses, crucial for global path planning.
• L (Landmarks): Semantic landmark information is extracted by Astra-Global from visual data at each node, enriching the map’s semantic understanding. These landmarks store semantic attributes and are connected to multiple nodes via co-visibility relationships.

In practical localization, Astra-Global’s self-localization and target localization capabilities leverage a coarse-to-fine two-stage process for visual-language localization. The coarse stage analyzes input images and localization prompts, detects landmarks, establishes correspondence with a pre-built landmark map, and filters candidates based on visual consistency. The fine stage then uses the query image and coarse output to sample reference map nodes from the offline map, comparing their visual and positional information to directly output the predicted pose.

For language-based target localization, the model interprets natural language instructions, identifies relevant landmarks using their functional descriptions within the map, and then leverages landmark-to-node association mechanisms to locate relevant nodes, retrieving target images and 6-DoF poses.

To empower Astra-Global with robust localization abilities, the team employed a meticulous training methodology. Using Qwen2.5-VL as the backbone, they combined Supervised Fine-Tuning (SFT) with Group Relative Policy Optimization (GRPO). SFT involved diverse datasets for various tasks, including coarse and fine localization, co-visibility detection, and motion trend estimation. In the GRPO phase, a rule-based reward function (including format, landmark extraction, map matching, and extra landmark rewards) was used to train for visual-language localization. Experiments showed GRPO significantly improved Astra-Global’s zero-shot generalization, achieving 99.9% localization accuracy in unseen home environments, surpassing SFT-only methods.

Astra-Local: The Intelligent Assistant for Local Planning

Astra-Local acts as the intelligent assistant for Astra’s high-frequency tasks, a multi-task network capable of efficiently generating local paths and accurately estimating odometry from sensor data. Its architecture comprises three core components: a 4D spatio-temporal encoder, a planning head, and an odometry head.

The 4D spatio-temporal encoder replaces traditional mobile stack perception and prediction modules. It begins with a 3D spatial encoder that processes N omnidirectional images through a Vision Transformer (ViT) and Lift-Splat-Shoot to convert 2D image features into 3D voxel features. This 3D encoder is trained using self-supervised learning via 3D volumetric differentiable neural rendering. The 4D spatio-temporal encoder then builds upon the 3D encoder, taking past voxel features and future timestamps as input to predict future voxel features through ResNet and DiT modules, providing current and future environmental representations for planning and odometry.

The planning head, based on pre-trained 4D features, robot speed, and task information, generates executable trajectories using Transformer-based flow matching. To prevent collisions, the planning head incorporates a masked ESDF loss (Euclidean Signed Distance Field). This loss calculates the ESDF of a 3D occupancy map and applies a 2D ground truth trajectory mask, significantly reducing collision rates. Experiments demonstrate its superior performance in collision rate and overall score on out-of-distribution (OOD) datasets compared to other methods.

The odometry head predicts the robot’s relative pose using current and past 4D features and additional sensor data (e.g., IMU, wheel data). It trains a Transformer model to fuse information from different sensors. Each sensor modality is processed by a specific tokenizer, combined with modality embeddings and temporal positional embeddi…

Konten dipersingkat otomatis.

🔗 Sumber: syncedreview.com