Search arXiv⌕ Search

arXiv subjects

Tim Jones

Publications and source records attributed to Tim Jones.

10 recordsLinked to original sources
arXiv2026

${\pi}_{0.7}$: a Steerable Generalist Robotic Foundation Model with Emergent Capabilities

We present a new robotic foundation model, called ${\pi}_{0.7}$, that can enable strong out-of-the-box performance in a wide range of scenarios. ${\pi}_{0.7}$ can follow diverse language instructions in unseen environments, including multi-stage tasks with various kitchen appliances, provide zero-shot cross-embodiment generalization, for example enabling a robot to fold laundry without seeing the task before, and perform challenging tasks such as operating an espresso machine out of the box at a level of performance that matches much more specialized RL-finetuned models. The main idea behind ${\pi}_{0.7}$ is to use diverse context conditioning during training. This conditioning information, contained in the prompt, makes it possible to steer the model precisely to perform many tasks with different strategies. It is conditioned not just on a language command that describes what it should do, but on additional multimodal information that also describes the manner or strategy in which it should do it, including metadata about task performance and subgoal images. This enables ${\pi}_{0.7}$ to use very diverse data, including demonstrations, potentially suboptimal (autonomous) data including failures, and data from non-robot sources. Our experiments evaluate ${\pi}_{0.7}$ across numerous tasks with multiple robot platforms, on tasks that require speed and dexterity, language following, and compositional task generalization.

cs.LG↗
arXiv2025

$\pi^{*}_{0.6}$: a VLA That Learns From Experience

We study how vision-language-action (VLA) models can improve through real-world deployments via reinforcement learning (RL). We present a general-purpose method, RL with Experience and Corrections via Advantage-conditioned Policies (RECAP), that provides for RL training of VLAs via advantage conditioning. Our method incorporates heterogeneous data into the self-improvement process, including demonstrations, data from on-policy collection, and expert teleoperated interventions provided during autonomous execution. RECAP starts by pre-training a generalist VLA with offline RL, which we call $\pi^{*}_{0.6}$, that can then be specialized to attain high performance on downstream tasks through on-robot data collection. We show that the $\pi^{*}_{0.6}$ model trained with the full RECAP method can fold laundry in real homes, reliably assemble boxes, and make espresso drinks using a professional espresso machine. On some of the hardest tasks, RECAP more than doubles task throughput and roughly halves the task failure rate.

cs.LG↗
arXiv2025

Silicon Wafer Fracture Stress for Tracking Sensors in Particle Physics Experiments

For the construction of the ATLAS Inner Tracker strip detector, silicon strip sensor modules are glued directly onto carbon fibre support structures using a soft silicone gel. During tests at temperatures below \unit[-35]{$^{\circ}$C}, several of the sensors were found to crack due to a mismatch in coefficients of thermal expansion between polyimide circuit boards with copper metal layers (glued onto the sensor) and the silicon sensor itself. While module assembly procedures were developed to minimise variations between modules, cold tests showed a wide range of temperatures at which supposedly comparable modules failed. The observed variance (fracture temperatures between \unit[-35]{\textcelsius} and \unit[-70]{\textcelsius}) for supposedly comparable modules suggests an undetected variation between modules suspected to be intrinsic to the silicon wafer itself. Therefore, a test programme was developed to investigate the fracture stress of representative sensor wafer cutoffs. This paper presents results for the fracture stress of silicon sensors used in detector modules.

physics.ins-det↗
arXiv2025

$\pi_{0.5}$: a Vision-Language-Action Model with Open-World Generalization

In order for robots to be useful, they must perform practically relevant tasks in the real world, outside of the lab. While vision-language-action (VLA) models have demonstrated impressive results for end-to-end robot control, it remains an open question how far such models can generalize in the wild. We describe $\pi_{0.5}$, a new model based on $\pi_{0}$ that uses co-training on heterogeneous tasks to enable broad generalization. $\pi_{0.5}$\ uses data from multiple robots, high-level semantic prediction, web data, and other sources to enable broadly generalizable real-world robotic manipulation. Our system uses a combination of co-training and hybrid multi-modal examples that combine image observations, language commands, object detections, semantic subtask prediction, and low-level actions. Our experiments show that this kind of knowledge transfer is essential for effective generalization, and we demonstrate for the first time that an end-to-end learning-enabled robotic system can perform long-horizon and dexterous manipulation skills, such as cleaning a kitchen or bedroom, in entirely new homes.

cs.LG↗
arXiv2024

$\pi_0$: A Vision-Language-Action Flow Model for General Robot Control

Robot learning holds tremendous promise to unlock the full potential of flexible, general, and dexterous robot systems, as well as to address some of the deepest questions in artificial intelligence. However, bringing robot learning to the level of generality required for effective real-world systems faces major obstacles in terms of data, generalization, and robustness. In this paper, we discuss how generalist robot policies (i.e., robot foundation models) can address these challenges, and how we can design effective generalist robot policies for complex and highly dexterous tasks. We propose a novel flow matching architecture built on top of a pre-trained vision-language model (VLM) to inherit Internet-scale semantic knowledge. We then discuss how this model can be trained on a large and diverse dataset from multiple dexterous robot platforms, including single-arm robots, dual-arm robots, and mobile manipulators. We evaluate our model in terms of its ability to perform tasks in zero shot after pre-training, follow language instructions from people and from a high-level VLM policy, and its ability to acquire new skills via fine-tuning. Our results cover a wide variety of tasks, such as laundry folding, table cleaning, and assembling boxes.

cs.LG↗
arXiv2021

Measuring the coefficient of moisture expansion of Hysol 9396 loaded with boron nitride powder

The future ATLAS ITk strip tracker will consist of 17,000 silicon strip detector modules mounted on support structures called cores. Cores are assembled from a number of components, among others carbon fibre facings, honeycomb structure and carbon foam surrounding titanium tubes used for cooling, using a two-component epoxy (Hysol 9396) loaded with boron nitride for good thermal conductivity. The adhesive constitutes about 20 % of a core's weight. During operation, the detector is cooled down to -40 $^\circ$C using bi-phase carbon dioxide and flushed with dry gas to prevent condensation. The effect of this temperature change has been simulated to study the impact of Coefficient of Thermal Expansion (CTE) mismatches between different materials and investigate resulting deformations and misalignment. In addition to the shrinking of an adhesive during cooling, which can be estimated well using its known CTE, flushing the detector volume with dry gas removes the moisture contained in the adhesive, leading to an additional shrinking. In order to estimate the impact of shrinking during drying, the Coefficient of Moisture Expansion (CME) of Hysol samples with different contents of boron nitride as well as their overall moisture absorption were measured and their extent compared to the contraction associated with cooling.

physics.ins-det↗
arXiv2019

Organization of machine learning based product development as per ISO 26262 and ISO/PAS 21448

Machine learning (ML) algorithms generate a continuous stream of success stories from various domains and enable many novel applications in safety-critical systems. With the advent of autonomous driving, ML algorithms are being used in the automotive domain, where the applicable functional safety standard is ISO 26262. However, requirements and recommendations provided by ISO 26262 do not cover specific properties of machine learning algorithms. Therefore, specific aspects of ML (e.g., dataset requirements, performance evaluation metrics, lack of interpretability) must be addressed within some work products, which collect documentation resulting from one or more associated requirements and recommendations of ISO 26262. In this paper, we propose how key technical aspects and supporting processes related to development of ML-based systems can be organized according to ISO 26262 phases, sub-phases, and work products. We follow the same approach as in the ISO/PAS 21448 standard, which complements ISO 26262, in order to account for edge cases that can lead to hazards not directly caused by system failure.%, but resulting from functional insufficiencies of the intended functionality or by reasonably foreseeable misuse by persons.

cs.LG↗
arXiv2015

Alternative glues for the production of ATLAS silicon strip modules for the Phase-II upgrade of the ATLAS Inner Detector

The Phase-II upgrade of the ATLAS detector for the High Luminosity Large Hadron Collider (HL-LHC) includes the replacement of the current Inner Detector with an all-silicon tracker consisting of pixel and strip detectors. The current Phase-II detector layout requires the construction of 20,000 strip detector modules consisting of sensor, circuit boards and readout chips, which are connected mechanically using adhesives. The adhesive between readout chips and circuit board is a silver epoxy glue as was used in the current ATLAS SemiConductor Tracker (SCT). This glue has several disadvantages, which motivated the search for an alternative. This paper presents a study concerning the use of six ultra-violet (UV) cure glues and a glue pad for use in the assembly of silicon strip detector modules for the ATLAS upgrade. Trials were carried out to determine the ease of use, the thermal conduction and shear strength, thermal cycling, radiation hardness, corrosion resistance and shear strength tests. These investigations led to the exclusion of three UV cure glues as well as the glue pad. Three UV cure glues were found to be possible better alternatives. Results from electrical tests of first prototype modules constructed using these glues are presented.

physics.ins-det↗
arXiv2015

Understanding the Heavy Tailed Dynamics in Human Behavior

The recent availability of electronic datasets containing large volumes of communication data has made it possible to study human behavior on a larger scale than ever before. From this, it has been discovered that across a diverse range of data sets, the inter-event times between consecutive communication events obey heavy tailed power law dynamics. Explaining this has proved controversial, and two distinct hypotheses have emerged. The first holds that these power laws are fundamental, and arise from the mechanisms such as priority queuing that humans use to schedule tasks. The second holds that they are a statistical artifact which only occur in aggregated data when features such as circadian rhythms and burstiness are ignored. We use a large social media data set to test these hypotheses, and find that although models that incorporate circadian rhythms and burstiness do explain part of the observed heavy tails, there is residual unexplained heavy tail behavior which suggests a more fundamental cause. Based on this, we develop a new quantitative model of human behavior which improves on existing approaches, and gives insight into the mechanisms underlying human interactions.

physics.soc-ph↗