MAROKO133 Breaking ai: ByteDance Introduces Astra: A Dual-Model Architecture for Autonomou

📌 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 Update ai: Professor in Epstein Files Makes Extremely Awkward Announce

It’s been a rocky week for public intellectual Larry Summers.

On Tuesday, the 70-year-old Harvard economics professor gave an awkward announcement to his class of college students acknowledging his embarrassing ties to the deceased sex trafficker and billionaire financier Jeffrey Epstein, which were brought to light with the recent release of another batch of Epstein’s emails.

“Some of you will have seen my statement of regret expressing my shame with respect to what I did in communication with Mr. Epstein, and that I’ve said that I’m going to step back from public activities, but for a time,” Summers intoned, gravely. 

“But I think it’s very important to fulfill my teaching obligations. And so, with your permission,” he continues, not waiting for anyone’s objections, “I’m gonna — we’re gonna go forward and, uh, talk about the material, uh, in the class.”

Summers’ preeminence as an economic authority has made him a mainstay of politics for decades, serving as Bill Clinton’s treasury secretary from 1999 to 2001, and an economic advisor under Barack Obama from 2009 to 2011. He was also president of Harvard University for five years, but stepped down in 2006 after being criticized for making sexist remarks about women.

His connection to Epstein was once again brought under the microscope after a House committee released a trove of Epstein’s emails and documents last week, in which it became clear that Summers was far closer to the deceased sex criminal than he had previously let on. On Monday, Summers released a statement expressing how “deeply ashamed” he was for communicating with Epstein, the spirit of which he repeated in his address to his class.

In a 2018 email exchange, Summers, who is married with three children, asked Epstein for romantic advice related to a woman he said he was a mentor for — and who was decidedly not his wife — while lamenting that he wouldn’t be seen as anything more than that.

“Think for now I’m going nowhere with her except economics mentor,” Summers wrote to Epstein. 

Epstein, referring to himself as Summers’ “wing man,” assured that the woman was “doomed to be with you.”

These exchanges, as did many others, took place well after Epstein had pleaded guilty in 2008 to sexually abusing teenage girls as young as 14 years old, and continued up until his arrest in 2019 for child sex trafficking.

This isn’t the first time Summers’ ties to Epstein have been exposed. In 2023, the Wall Street Journal revealed that the pair met more than a dozen times between 2013 and 2016, during which Summers asked Epstein’s advice on raising $1 million for his wife’s poetry project. Epstein later chipped in $110,000, via a nonprofit.

Not much came of those revelations, but this time, it doesn’t seem Summers will be let off the hook quite so easily. On Wednesday, he announced he was stepping down from his position on OpenAI’s board of directors, conceding influence in what is, for better or worse, one of the most important companies in the world right now.

It also appears he won’t be allowed to finish teaching his class, as he told his students he would. Instead, according to his spokesperson, “co-teachers will complete the remaining three class sessions of the courses he has been teaching with them this semester, and he is not scheduled to teach next semester.”

More on Epstein: New Evidence Links Elon Musk to Epstein’s Island

The post Professor in Epstein Files Makes Extremely Awkward Announcement at Start of Class appeared first on Futurism.

🔗 Sumber: futurism.com