MAROKO133 Hot ai: ByteDance Introduces Astra: A Dual-Model Architecture for Autonomous Rob

📌 MAROKO133 Hot ai: ByteDance Introduces Astra: A Dual-Model Architecture for Auto

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

📌 MAROKO133 Breaking ai: Scientists Working on "Smart Dust" That Can Spy

In his 1963 scifi story “The Invincible,” the Polish writer Stanisław Lem imagined an artificial species of free-floating nanobots which roamed the atmosphere of a far-off planet. Like tiny bugs, the microscopic beings were powerless alone, but together they could form cooperative swarms to gather energy, reproduce, and ultimately defend their territory from predators with deadly force.

Unlike the story’s human protagonists, the “black cloud” of bots was incapable of reasoning beyond the simple logic of animal instincts. But when the two life forms inevitably come into conflict, literary critic Jerzy Jarzębski writes, human evolution proves its mettle over the mindless automaton — not by eradicating the deadly species, but by making a conscious decision to let it live.

Lem probably never imagined his evolutionary parable of living dust was just a few decades from becoming a reality — or that it would become the inspiration for the development of a real-life military technology known as “smart dust.”

Starting out as a theoretical research proposal to the Defense Advanced Research Projects Agency (DARPA) — the Cold War-era military tech bureau behind everything from GPS navigation to the modern internet — smart dust is now being developed for use in a wide variety of industries, from environmental studies to commercial mining.

That’s according to Interesting Engineering, which recently published a rundown of the state of present-day smart dust after decades of development. Though “dust” remains a bit of a misnomer — it’s more like a bunch of tiny sensors capable of delaying data to a central device — there’s a large body of theoretical and simulated work laying a path for practical microengineering that’s steadily coming into its own.

Case in point, while nanotech began as an effort to build relatively simple wireless receivers around the size of a grain of rice, thanks to decades of R&D, some motes being developed are now nearly invisible to the naked eye, measuring in at anywhere from 1 cubic millimeter to .02 cubic millimeters.

As early as 2003, micro-sensor platforms like Crossbow Technology, Inc’s “MICA” and UC Berkeley’s “Spec” have successfully detected all kinds of variables while measuring in at mere millimeters, recording changes in humidity, light and temperature.

Recent developments within the past 10 years have expanded these sensor’s abilities to record various levels of sound, and work is underway to develop motes capable of detecting the chemical composition of the air. They can be used individually to record changes in the human body, or deployed in swarms to identify biological compounds.

In the future, the mites are hoped to be able to report a near-infinite amount of data in suspended, 3D environments — like a microscopic version of Bill Paxton’s “Dorothy” sensors in the 1996 weather thriller “Twister.”

Per IE, the current “smart dust industry,” made up of tech companies like Emerson Process Management and Hewlett-Packard, was valued at around $115 million in 2022. By 2032, it’s expected to reach nearly $400 million.

While various militaries are keen on developing smart dust for intelligence reasons, much of the present research is carried in university and corporate labs. An Israeli firm called Stardust Solutions, for example, drew concerns from the Bulletin of the Atomic Scientists when it announced its intentions of releasing a variation on smart dust to block out the Sun — involving inert particular matter in conjunction with an atmospheric monitoring system — in violation of international geoengineering laws.

While the tech is pretty dystopian as it is, there’s a lot of room for improvement. The need to interface with a centralized data-processing unit, for example, means the tiny units can’t travel too far from their human controller. Their usable lifespan is likewise pretty short, though that’s changing with innovations in energy-harvesting capabilities via light, vibrations, and electromagnetic fields.

One thing’s abundantly clear: now might be a good time to invest in an air purifier.

More on tiny tech: Chinese Military Shows Off Fly-Sized Drones for Covert Ops

The post Scientists Working on "Smart Dust" That Can Spy on a Room While Drifting Throught the Air appeared first on Futurism.

🔗 Sumber: futurism.com