Software : Cloud Computing : General
Our mission is to organize the world’s information and make it universally accessible and useful.
RT-1: Robotics Transformer for Real-World Control at Scale
Major recent advances in multiple subfields of machine learning (ML) research, such as computer vision and natural language processing, have been enabled by a shared common approach that leverages large, diverse datasets and expressive models that can absorb all of the data effectively. Although there have been various attempts to apply this approach to robotics, robots have not yet leveraged highly-capable models as well as other subfields.
Several factors contribute to this challenge. First, there’s the lack of large-scale and diverse robotic data, which limits a model’s ability to absorb a broad set of robotic experiences. Data collection is particularly expensive and challenging for robotics because dataset curation requires engineering-heavy autonomous operation, or demonstrations collected using human teleoperations. A second factor is the lack of expressive, scalable, and fast-enough-for-real-time-inference models that can learn from such datasets and generalize effectively.
To address these challenges, we propose the Robotics Transformer 1 (RT-1), a multi-task model that tokenizes robot inputs and outputs actions (e.g., camera images, task instructions, and motor commands) to enable efficient inference at runtime, which makes real-time control feasible. This model is trained on a large-scale, real-world robotics dataset of 130k episodes that cover 700+ tasks, collected using a fleet of 13 robots from Everyday Robots (EDR) over 17 months. We demonstrate that RT-1 can exhibit significantly improved zero-shot generalization to new tasks, environments and objects compared to prior techniques. Moreover, we carefully evaluate and ablate many of the design choices in the model and training set, analyzing the effects of tokenization, action representation, and dataset composition. Finally, we’re open-sourcing the RT-1 code, and hope it will provide a valuable resource for future research on scaling up robot learning.
Using AI to increase asset utilization and production uptime for manufacturers
Google Cloud created purpose-built tools and solutions to organize manufacturing data, make it accessible and useful, and help manufacturers to quickly take significant steps on this journey by reducing the time to value. In this post, we will explore a practical example of how manufacturers can use Google Cloud manufacturing solutions to train, deploy and extract value from ML-enabled capabilities to predict asset utilization and maintenance needs. The first step to a successful machine learning project is to unify necessary data in a common repository. For this, we will use Manufacturing Connect, the factory edge platform co-developed with Litmus Automation, to connect to manufacturing assets and stream the asset telemetries to Pub/Sub.
The following scenario is based on a hypothetical company, Cymbal Materials. This company is a factitious discrete manufacturing company that runs 50+ factories in 10+ countries. 90% of Cymbal Materials manufacturing processes involve milling, which are accomplished using industrial computer numerical control (CNC) milling machines. Although their factories implement routine maintenance checklists, there are unplanned and unknown failures that happen occasionally. However, many of the Cymbal Materials factory workers lack the experience to identify and troubleshoot failures due to labor shortage and high turnover rate in their factories. Hence, Cymbal Materials is working with Google Cloud to build a machine learning model that can identify and analyze failures on top of Manufacturing Connect, Manufacturing Data Engine, and Vertex AI.
Intro to deep learning to track deforestation in supply chains
In my experience, I have observed that it’s common in machine learning to surrender to the process of experimenting with many different algorithms in a trial and error fashion, until you get the desired result. My peers and I at Google have a People and Planet AI YouTube series where we talk about how to train and host a model for environmental purposes using Google Cloud and Google Earth Engine. Our focus is inspiring people to use deep learning, and if we could rename the series, we would call it AI for Minimalists since we would recommend artificial neural networks for most of our use cases. And so in this episode we give an overview of what deep learning is and how you can use it for tracking deforestation in supply chains.
The art of effective factory data visualization
Table Tennis: A Research Platform for Agile Robotics
Robot learning has been applied to a wide range of challenging real world tasks, including dexterous manipulation, legged locomotion, and grasping. It is less common to see robot learning applied to dynamic, high-acceleration tasks requiring tight-loop human-robot interactions, such as table tennis. There are two complementary properties of the table tennis task that make it interesting for robotic learning research. First, the task requires both speed and precision, which puts significant demands on a learning algorithm. At the same time, the problem is highly-structured (with a fixed, predictable environment) and naturally multi-agent (the robot can play with humans or another robot), making it a desirable testbed to investigate questions about human-robot interaction and reinforcement learning. These properties have led to several research groups developing table tennis research platforms.
How Boeing overcame their on-premises implementation challenges with data & AI
CircularNet: Reducing waste with Machine Learning
The facilities where our waste and recyclables are processed are called “Material Recovery Facilities” (MRFs). Each MRF processes tens of thousands of pounds of our societal “waste” every day, separating valuable recyclable materials like metals and plastics from non-recyclable materials. A key inefficiency within the current waste capture and sorting process is the inability to identify and segregate waste into high quality material streams. The accuracy of the sorting directly determines the quality of the recycled material; for high-quality, commercially viable recycling, the contamination levels need to be low. Even though the MRFs use various technologies alongside manual labor to separate materials into distinct and clean streams, the exceptionally cluttered and contaminated nature of the waste stream makes automated waste detection challenging to achieve, and the recycling rates and the profit margins stay at undesirably low levels.
Enter what we call “CircularNet”, a set of models that lowers barriers to AI/ML tech for waste identification and all the benefits this new level of transparency can offer. Our goal with CircularNet is to develop a robust and data-efficient model for waste/recyclables detection, which can support the way we identify, sort, manage, and recycle materials across the waste management ecosystem.
Synopsys helps semiconductor designers accelerate chip design and development on Google Cloud
EDA software is a large consumer of high performance computing capacity in the cloud. With the release of Synopsys Cloud bring-your-own-cloud (BYOC) solution on Google Cloud, chip designers can now scale their Google Cloud infrastructure with Synopsys’s leading EDA tools under the flexible FlexEDA pay-per-use model and access unlimited EDA software license availability on-demand by the hour or minute.
Lufthansa increases on-time flights by wind forecasting with Google Cloud ML
The magnitude and direction of wind significantly impacts airport operations, and Lufthansa Group Airlines are no exception. A particularly troublesome kind is called BISE: it is a cold, dry wind that blows from the northeast to southwest in Switzerland, through the Swiss Plateau. Its effects on flight schedules can be severe, such as forcing planes to change runways, which can create a chain reaction of flight delays and possible cancellations. In Zurich Airport, in particular, BISE can potentially reduce capacity by up to 30%, leading to further flight delays and cancellations, and to millions in lost revenue for Lufthansa (as well as dissatisfaction among their passengers).
Machine learning (ML) can help airports and airlines to better anticipate and manage these types of disruptive weather events. In this blog post, we’ll explore an experiment Lufthansa did together with Google Cloud and its Vertex AI Forecast service, accurately predicting BISE hours in advance, with more than 40% relative improvement in accuracy over internal heuristics, all within days instead of the months it often takes to do ML projects of this magnitude and performance.
How Volkswagen and Google Cloud are using machine learning to design more energy-efficient cars
Volkswagen strives to design beautiful, performant, and energy efficient vehicles. This entails an iterative process where designers go through many design drafts, evaluating each, integrating the feedback, and refining. For example, a vehicle’s drag coefficient—its resistance to air—is one of the most important factors of energy efficiency. Thus, getting estimates of the drag coefficient for several designs helps the designers experiment and converge toward more energy-efficient solutions. The cheaper and faster this feedback loop is, the more it enables the designers.
This joint research effort between Volkswagen and Google has produced promising results with the help of the Vertex AI platform. In this first milestone, the team was able to successfully bring recent AI research results a step closer to practical application for car design. This first iteration of the algorithm can produce a drag coefficient estimate with an average error of just 4%, within a second. An average error of 4%, while not quite as accurate as a physical wind tunnel test, can be used to narrow a large selection of design candidates to a small shortlist. And given how quickly the estimates appear, we have made a substantial improvement on the existing methods that take days or weeks. With the algorithm that we have developed, designers can run more efficiency tests, submit more candidates, and iterate towards richer, more effective designs in just a small fraction of the time previously required.
Towards Helpful Robots: Grounding Language in Robotic Affordances
In “Do As I Can, Not As I Say: Grounding Language in Robotic Affordances”, we present a novel approach, developed in partnership with Everyday Robots, that leverages advanced language model knowledge to enable a physical agent, such as a robot, to follow high-level textual instructions for physically-grounded tasks, while grounding the language model in tasks that are feasible within a specific real-world context. We evaluate our method, which we call PaLM-SayCan, by placing robots in a real kitchen setting and giving them tasks expressed in natural language. We observe highly interpretable results for temporally-extended complex and abstract tasks, like “I just worked out, please bring me a snack and a drink to recover.” Specifically, we demonstrate that grounding the language model in the real world nearly halves errors over non-grounded baselines. We are also excited to release a robot simulation setup where the research community can test this approach.
Aramco and Cognite join forces in new data venture
Aramco and Cognite, a global leader in industrial software, have launched CNTXT, a joint venture based in the Kingdom of Saudi Arabia. Headquartered in Riyadh, CNTXT aims to support the Kingdom’s industrial digitalization, and the wider MENA region.
CNTXT will provide digital transformation services enabled by advanced cloud solutions and leading industrial software. These solutions and services aim to help public and private sector companies to future-proof their data infrastructure, increase revenue, cut costs and reduce risks while enhancing operational sustainability and security. CNTXT is Google Cloud’s reseller for cloud solutions in the Kingdom and the exclusive reseller of Cognite Data Fusion in MENA region. Additionally, Google Cloud is expected to launch a “Center of Excellence” later this year to provide training to developers and business leaders in how to use cloud technologies.
TELUS: Solving for workers’ safety with edge computing and 5G
Together with Google Cloud, we have been leveraging solutions with the power of MEC and 5G to develop a workers’ safety application in our Edmonton Data Center that enables on-premise video analytics cameras to screen manufacturing facilities and ensure compliance with safety requirements to operate heavy-duty machinery. The CCTV (closed-circuit television) cameras we used are cost-effective and easier to deploy than RTLS (real time location services) solutions that detect worker proximity and avoid collisions. This is a positive, proactive step to steadily improve workplace safety. For example, if a worker’s hand is close to a drill, that drill press will not bore holes in any surface until the video analytics camera detects that the worker’s hand has been removed from the safety zone area.
Introducing new Google Cloud manufacturing solutions: smart factories, smarter workers
The new manufacturing solutions from Google Cloud give manufacturing engineers and plant managers access to unified and contextualized data from across their disparate assets and processes.
Manufacturing Data Engine is the foundational cloud solution to process, contextualize and store factory data. The cloud platform can acquire data from any type of machine, supporting a wide range of data, from telemetry to image data, via a private, secure, and low cost connection between edge and cloud. With built-in data normalization and context-enrichment capabilities, it provides a common data model, with a factory-optimized data lakehouse for storage.
Manufacturing Connect is the factory edge platform co-developed with Litmus Automation that quickly connects with nearly any manufacturing asset via an extensive library of 250-plus machine protocols. It translates machine data into a digestible dataset and sends it to the Manufacturing Data Engine for processing, contextualization and storage. By supporting containerized workloads, it allows manufacturers to run low-latency data visualization, analytics and ML capabilities directly on the edge.
Price optimization notebook for apparel retail using Google Vertex AI
One of the key requirements of a price optimization system is an accurate forecasting model to quickly simulate demand response to price changes. Historically, developing a Machine Learning forecast model required a long timeline with heavy involvement from skilled specialists in data engineering, data science, and MLOps. The teams needed to perform a variety of tasks in feature engineering, model architecture selection, hyperparameter optimization, and then manage and monitor deployed models.
Vertex AI Forecast provides advanced AutoML workflow for time series forecasting which helps dramatically reduce the engineering and research effort required to develop accurate forecasting models. The service easily scales up to large datasets with over 100 million rows and 1000 columns, covering years of data for thousands of products with hundreds of possible demand drivers. Most importantly it produces highly accurate forecasts. The model scored in the top 2.5% of submissions in M5, the most recent global forecasting competition which used data from Walmart.
UPS Expands Deal With Google Cloud to Prepare for Surge in Data
Logistics company to gain network, storage and compute capacity to help it analyze new data coming from initiatives such as RFID chips on packages
Robust Routing Using Electrical Flows
We view the road network as a graph, where intersections are nodes and roads are edges. Our method then models the graph as an electrical circuit by replacing the edges with resistors, whose resistances equal the road traversal time, and then connecting a battery to the origin and destination, which results in electrical current between those two points. In this analogy, the resistance models how time-consuming it is to traverse a segment. In this sense, long and congested segments have high resistances.
Improving PPA In Complex Designs With AI
The goal of chip design always has been to optimize power, performance, and area (PPA), but results can vary greatly even with the best tools and highly experienced engineering teams. AI works best in design when the problem is clearly defined in a way that AI can understand. So an IC designer must first see if there is a problem that can be tied to a system’s ability to adapt to, learn, and generalize knowledge/rules, and then apply these knowledge/rules to an unfamiliar scenario.
Can Robots Follow Instructions for New Tasks?
The results of this research show that simple imitation learning approaches can be scaled in a way that enables zero-shot generalization to new tasks. That is, it shows one of the first indications of robots being able to successfully carry out behaviors that were not in the training data. Interestingly, language embeddings pre-trained on ungrounded language corpora make for excellent task conditioners. We demonstrated that natural language models can not only provide a flexible input interface to robots, but that pretrained language representations actually confer new generalization capabilities to the downstream policy, such as composing unseen object pairs together.
In the course of building this system, we confirmed that periodic human interventions are a simple but important technique for achieving good performance. While there is a substantial amount of work to be done in the future, we believe that the zero-shot generalization capabilities of BC-Z are an important advancement towards increasing the generality of robotic learning systems and allowing people to command robots. We have released the teleoperated demonstrations used to train the policy in this paper, which we hope will provide researchers with a valuable resource for future multi-task robotic learning research.
Inside X’s Mission to Make Robots Boring
It’s research by Everyday Robots, a project of X, Alphabet’s self-styled “moonshot factory.” The cafe testing ground is one of dozens on the Google campus in Mountain View, California, where a small percentage of the company’s massive workforce has now returned to work. The project hopes to make robots useful, operating in the wild instead of controlled environments like factories. After years of development, Everyday Robots is finally sending its robots into the world—or at least out of the X headquarters building—to do actual work.
Chip floorplanning with deep reinforcement learning
AWS, Google, Microsoft apply expertise in data, software to manufacturing
As manufacturing becomes digitized, Google’s methodologies that were developed for the consumer market are becoming relevant for industry, said Wee, who previously worked in the semiconductor industry as an industrial engineer. “We believe we’re at a point in time where these technologies—primarily the analytics and AI area—that have been very difficult to use for the typical industrial engineer are becoming so easy to use on the shop floor,” he said. “That’s where we believe our competitive differentiation lies.”
Meanwhile, Ford is also selectively favoring human brain power over software to analyze data and turning more and more to in-house coders than applications vendors. “The solution will be dependent upon the application,” Mikula said. “Sometimes it will be software, and sometimes it’ll be a data analyst who crunches the data sources. We would like to move to solutions that are more autonomous and driven by machine learning and artificial intelligence. The goal is to be less reliant on purchased SaaS.”
Intrinsic is working to unlock the creative and economic potential of industrial robotics for millions more businesses, entrepreneurs, and developers. We’re developing software tools designed to make industrial robots (which are used to make everything from solar panels to cars) easier to use, less costly and more flexible, so that more people can use them to make new products, businesses and services.
Visual Inspection AI: a purpose-built solution for faster, more accurate quality control
The Google Cloud Visual Inspection AI solution automates visual inspection tasks using a set of AI and computer vision technologies that enable manufacturers to transform quality control processes by automatically detecting product defects.
We built Visual Inspection AI to meet the needs of quality, test, manufacturing, and process engineers who are experts in their domain, but not in AI. By combining ease of use with a focus on priority uses cases, customers are realizing significant benefits compared to general purpose machine learning (ML) approaches.
Toward Generalized Sim-to-Real Transfer for Robot Learning
A limitation for their use in sim-to-real transfer, however, is that because GANs translate images at the pixel-level, multi-pixel features or structures that are necessary for robot task learning may be arbitrarily modified or even removed.
To address the above limitation, and in collaboration with the Everyday Robot Project at X, we introduce two works, RL-CycleGAN and RetinaGAN, that train GANs with robot-specific consistencies — so that they do not arbitrarily modify visual features that are specifically necessary for robot task learning — and thus bridge the visual discrepancy between sim and real.
Learning to Manipulate Deformable Objects
In “Learning to Rearrange Deformable Cables, Fabrics, and Bags with Goal-Conditioned Transporter Networks,” to appear at ICRA 2021, we release an open-source simulated benchmark, called DeformableRavens, with the goal of accelerating research into deformable object manipulation. DeformableRavens features 12 tasks that involve manipulating cables, fabrics, and bags and includes a set of model architectures for manipulating deformable objects towards desired goal configurations, specified with images. These architectures enable a robot to rearrange cables to match a target shape, to smooth a fabric to a target zone, and to insert an item in a bag. To our knowledge, this is the first simulator that includes a task in which a robot must use a bag to contain other items, which presents key challenges in enabling a robot to learn more complex relative spatial relations.
Google Cloud and Seagate: Transforming hard-disk drive maintenance with predictive ML
At Google Cloud, we know first-hand how critical it is to manage HDDs in operations and preemptively identify potential failures. We are responsible for running some of the largest data centers in the world—any misses in identifying these failures at the right time can potentially cause serious outages across our many products and services. In the past, when a disk was flagged for a problem, the main option was to repair the problem on site using software. But this procedure was expensive and time-consuming. It required draining the data from the drive, isolating the drive, running diagnostics, and then re-introducing it to traffic.
That’s why we teamed up with Seagate, our HDD original equipment manufacturer (OEM) partner for Google’s data centers, to find a way to predict frequent HDD problems. Together, we developed a machine learning (ML) system, built on top of Google Cloud, to forecast the probability of a recurring failing disk—a disk that fails or has experienced three or more problems in 30 days.
Multi-Task Robotic Reinforcement Learning at Scale
For general-purpose robots to be most useful, they would need to be able to perform a range of tasks, such as cleaning, maintenance and delivery. But training even a single task (e.g., grasping) using offline reinforcement learning (RL), a trial and error learning method where the agent uses training previously collected data, can take thousands of robot-hours, in addition to the significant engineering needed to enable autonomous operation of a large-scale robotic system. Thus, the computational costs of building general-purpose everyday robots using current robot learning methods becomes prohibitive as the number of tasks grows.
Way beyond AlphaZero: Berkeley and Google work shows robotics may be the deepest machine learning of all
With no well-specified rewards and state transitions that take place in a myriad of ways, training a robot via reinforcement learning represents perhaps the most complex arena for machine learning.
Rearranging the Visual World
Transporter Nets use a novel approach to 3D spatial understanding that avoids reliance on object-centric representations, making them general for vision-based manipulation but far more sample efficient than benchmarked end-to-end alternatives. As a consequence, they are fast and practical to train on real robots. We are also releasing an accompanying open-source implementation of Transporter Nets together with Ravens, our new simulated benchmark suite of ten vision-based manipulation tasks.
Edge-Inference Architectures Proliferate
What makes one AI system better than another depends on a lot of different factors, including some that aren’t entirely clear.
The new offerings exhibit a wide range of structure, technology, and optimization goals. All must be gentle on power, but some target wired devices while others target battery-powered devices, giving different power/performance targets. While no single architecture is expected to solve every problem, the industry is in a phase of proliferation, not consolidation. It will be a while before the dust settles on the preferred architectures.
RightHand Robotics raises $23 million from Menlo Ventures, Google
With its reinforced bank account, Somerville, Mass.-based RightHand plans to expand its business and technical teams and broaden its suite of product applications, the firm said. “The funds will be used to support our growth and in hiring people as fast as we effectively can,” Martinelli said. “We’re getting follow-on orders and we need to support those orders and extend the product line, both for projects in the U.S. and in Europe and Japan.”
Google Glass Didn't Disappear. You Can Find It On The Factory Floor
With Google Glass, she scans the serial number on the part she’s working on. This brings up manuals, photos or videos she may need. She can tap the side of headset or say “OK Glass” and use voice commands to leave notes for the next shift worker.
Peggy Gullick, business process improvement director with AGCO, says the addition of Google Glass has been “a total game changer.” Quality checks are now 20 percent faster, she says, and it’s also helpful for on-the-job training of new employees. Before this, workers used tablets.
Augmented Reality Is Already Improving Worker Performance
The video below, for example, shows a side-by-side time-lapse comparison of a GE technician wiring a wind turbine’s control box using the company’s current process, and then doing the same task while guided by line-of-sight instructions overlaid on the job by an AR headset. The device improved the worker’s performance by 34% on first use.
There’s been concern about machines replacing human workers, and certainly this is happening for some jobs. But the experience at General Electric and other industrial firms shows that, for many jobs, combinations of humans and machines outperform either working alone. Wearable augmented reality devices are especially powerful, as they deliver the right information at the right moment and in the ideal format, directly in workers’ line of sight, while leaving workers’ hands free so they can work without interruption. This dramatically reduces the time needed to complete a job because workers needn’t stop what they’re doing to flip through a paper manual or engage with a device or workstation. It also reduces errors because the AR display provides explicit guidance overlaid on the work being done, delivered on demand. Workers need only follow the detailed instructions directly in front of them in order to move through a sequence of steps to completion. If they encounter problems, they can launch training videos or connect by video with remote experts to share what they see through their smart glasses and get real-time assistance.