Scientists tracked kids for 8 years. What they found about screen time was surprising
Robot swarms trained by trial and error in simulation keep working on real robots
Reinforcement learning lets robots learn behaviors on their own, through trial and error, usually inside a computer simulation. Moving what they learn onto physical robots is another matter. Simulated and real robots never behave quite the same way, a mismatch known as the “reality gap,” and reinforcement learning has long been particularly sensitive to it. Researchers at IRIDIA, the artificial intelligence laboratory of the Université libre de Bruxelles (ULB), have shown how to get around it for robot swarms: swarms trained with their method in simulation keep performing reliably once they are moved onto real robots.
Teaching a swarm of robots to search for survivors after an earthquake, or to explore a hostile site, entirely inside a computer before risking a real machine sounds like the sensible way to do it. The catch is the reality gap. Friction, sensor noise and mechanical imprecision rarely behave the way they do in a simulation, and skills learned there tend to fall apart when a robot meets the real world.
Ilyes Gharbi, Mauro Birattari and colleagues at IRIDIA (Brussels School of Engineering, ULB) report their results in Science Robotics.
Reinforcement learning has become one of the most powerful tools in robotics for producing complex behaviors automatically, and it is behind some of the most striking recent advances in legged locomotion and robotic manipulation. Swarm robotics followed a different path. Its own tradition of modular design methods assembles robot behaviors from predefined building blocks. These methods transfer reliably to real robots, but they rely on slow, costly search procedures. The two lines of research had rarely been compared on solid experimental ground.
The researchers’ idea was to give the robots less freedom in exchange for more reliability. Instead of letting the learning algorithm set each wheel’s speed directly, they had it choose, moment by moment, among a handful of tried-and-tested behaviors: moving toward a light, aggregating with peers, exploring. This restriction keeps the robots from learning to exploit quirks of the simulator that don’t exist in reality. What is left to learn is the part that matters: when to switch from one behavior to another so that the group works together.
“For years we have argued that the reality gap is, at its core, a problem of overfitting: a design method that is too flexible learns the simulator rather than the task,” says Mauro Birattari, who led the research. “Reinforcement learning is very good at finding whatever works in simulation, including tricks that only work there. By limiting what the robots can learn, we leave the algorithm only the decisions that still make sense on a real robot.”
The team tested the approach on a swarm of 20 e-puck robots across five different collective missions, with 250 experiments on the physical robots in total. Swarms trained this way transferred to reality reliably. They substantially outperformed a state-of-the-art method in which the algorithm controls the robots’ motors directly, and which struggled once it left the simulator. The researchers also found that simpler, more constrained policies transferred better than larger, more flexible ones. That goes against a widespread intuition in artificial intelligence, that bigger and more flexible models are generally better.
“What excites me is that this makes reinforcement learning a practical tool for swarm robotics,” says Ilyes Gharbi, first author of the study. “Designing collective behaviors by hand takes experts a lot of time. With this approach, a swarm can be trained in simulation for a new mission and deployed on real robots with confidence that it will behave as expected. The next step is to test it on larger swarms, other robot platforms and less controlled environments.”
The study brings together two research traditions that had developed largely separately: the efficiency of reinforcement learning and the reliability of swarm robotics ’modular methods. Both will be needed before groups of autonomous robots can be trusted with tasks such as search and rescue, environmental monitoring, or exploring places too dangerous for people.
This research was partially supported by the Belgian Fonds de la Recherche Scientifique—FNRS through the PDR project “Reinforcement Learning for Robot Swarms.”
Publication
I. Gharbi, M. Kegeleirs, G. Legarda Herranz, J. Szpirer, K. Hasselmann, M. Birattari, “Structural bias enables reliable sim-to-real transfer in reinforcement learning for robot swarms,” Science Robotics (2026).
Quantum computers are getting too powerful to check. Scientists have a solution
A shared signal for humans and robots: Flexible display uses both color and near-infrared to communicate
Self-illuminating fibers reveal soft robot movements without batteries or complex electronic circuits
From AI Assistance to Production Decisions in Manufacturing
Watch the IROS 2026 plenary and keynote talks
The 2026 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) was held in Pittsburgh from 27 September to 1 October. The event brought together researchers and industry professionals with expertise spanning many different aspects of robotics.
IEEE have made the keynote and plenary panel recordings available to watch. You can also catch the opening and awards ceremonies, and the town hall meeting.
Watch IEEE.tv here.
Available to watch are the following:
- Plenary panels
- Start-ups and entrepreneurship
- Space
- Should robots be generalists or specialists?
- Keynote talks
- Elena De Momi – Open questions on the road to autonomous surgery
- Naomi Fitter – Socially-informed context awareness for robots in human-populated settings
- Greg Chirikjian – From hyper-redundancy to robot imagination and affordance learning
- Al Rizzi – Developing advanced robot capability at the RAI Institute
- Gaurav Bansal – Building autonomy for the real world
- Genta Suzuki – Fujitsu Kozuchi Physical OS: a research strategy for scalable physical AI
- Yigit Menguc – Soft robotics, tensegrity, and the people who love them
- Robin Murphy – Rethinking disaster robotics: the open questions that matter now
- Jen Jen Chung – All data are equal, but some data are more equal than others
- Tim Bretl – How you can be part of a revolution in composite manufacturing
- Lael Odhner – Big bets in manipulation systems: where is the frontier today?
- Gina Olson – Complexity in soft and bio-inspired robots
- David Held – Robots that learn after deployment: designing policies for fast adaptation
- Brittany Duncan – Patterns from the field: how contextual lessons can inform robot design
- Ankur Mehta – Robotics problems that robotics can’t solve
ERP Testing in a Connected Enterprise
ERP Testing in a Connected Enterprise: How to Maintain Quality Across Business Processes?
Your ERP system connects some of the most important functions of your business, from finance and procurement to inventory, human resources, supply chain, and reporting. When these functions work together, your teams gain better visibility, more consistent information, and greater control over business operations.
However, maintaining that reliability becomes increasingly challenging as your ERP environment evolves.
Cloud adoption, application upgrades, third-party integrations, custom configurations, and changing business requirements introduce new dependencies. A seemingly minor change in one module can affect transactions or workflows across several departments. Without adequate validation, these issues may reach production and disrupt everyday operations.
This makes ERP testing an ongoing business requirement rather than an activity limited to implementation or major upgrades. You need a testing approach that validates individual functions, connected applications, and complete business processes as your technology landscape changes.
What is ERP testing and why is it important?
ERP testing is the process of validating enterprise resource planning applications to ensure that modules, business workflows, integrations, and data function correctly. It helps organizations identify defects, prevent data inconsistencies, and maintain reliable business operations during implementations, upgrades, migrations, and ongoing changes.
Why Is ERP Testing Becoming More Challenging?
The complexity of ERP testing has increased alongside the way organizations use enterprise applications. Your ERP environment may include multiple modules, customized workflows, cloud platforms, legacy applications, and external systems that exchange information continuously.
Each additional connection creates dependencies that need to be considered during testing.
- Growing integration dependencies
Your ERP platform rarely operates in isolation. It may exchange information with CRM, HRMS, payment gateways, warehouse management systems, supplier portals, and business intelligence applications.
When an integration changes, you need to verify more than whether the connection works. You also need to confirm that information is transferred accurately, business rules remain intact, and dependent workflows continue without interruption.
- Frequent updates and application changes
Cloud ERP platforms introduce regular updates, patches, and feature releases. Your organization may also introduce configuration changes, new modules, or customized functionality to support evolving business requirements.
Every change creates a need for validation. Without an effective regression testing strategy, you risk introducing defects into workflows that previously operated as expected.
- Data inconsistencies across modules
Your business decisions depend on accurate ERP data. Incorrect field mappings, incomplete migration, duplicate records, and inconsistent information can affect financial reporting, inventory visibility, procurement, and operational planning.
You need to validate data at different stages of processing, including migration, integration, transaction execution, and reporting.
- Greater business impact from application defects
An ERP defect can affect multiple departments simultaneously. An incorrect calculation may influence financial records, while an inventory synchronization issue can affect order fulfillment and procurement planning.
The longer these defects remain undetected, the more difficult they can become to identify and resolve. Testing helps you uncover potential issues before they affect business-critical operations.
Why Should You Test ERP Workflows End to End?
Testing individual ERP modules is important, but it does not always provide a complete picture of application quality.
Consider a typical procure-to-pay process. It starts with a purchase requisition, moves through approval and purchase order creation, and continues through goods receipt, invoice processing, and payment.
Each stage may involve different ERP modules, users, approval rules, and connected applications.
Even if procurement, inventory, and finance modules pass their individual functional tests, an issue in the information exchange between them could interrupt the overall process.
End-to-end ERP testing helps you validate how these components work together.
For example, when testing your procure-to-pay workflow, you need to confirm that:
- Purchase requisitions follow the correct approval rules.
- Purchase orders contain accurate supplier, pricing, and quantity information.
- Goods receipts update inventory records correctly.
- Invoice matching and calculations follow defined business rules.
- Payment transactions generate the appropriate financial entries.
- Reports and reconciliation records reflect the completed transaction accurately.
You should also validate exception scenarios, such as rejected approvals, partial deliveries, duplicate invoices, and failed transactions.
This gives you greater confidence that your ERP system supports the complete business process, including situations that do not follow the standard transaction path.
What Should Your ERP Testing Strategy Cover?
Your testing strategy should reflect how your organization uses its ERP environment, the risks associated with application changes, and the business processes that require the highest level of reliability.
Depending on your ERP implementation and operational requirements, your testing scope may include:
- Functional testing: Verify that ERP modules, configurations, calculations, workflows, and business rules perform according to documented requirements.
- Integration testing: Validate information exchange between ERP modules and connected applications, including CRM, HRMS, supply chain, payment, and third-party systems.
- Regression testing: Revalidate existing functionality after upgrades, patches, configuration changes, and application enhancements to identify unintended effects.
- Data migration testing: Check migrated records for completeness, accuracy, consistency, field mapping, and reconciliation against source data.
- Performance testing: Assess application responsiveness and stability under expected workloads and transaction volumes, particularly during business-critical operations.
- Security and access testing: Validate role-based permissions, user access, and authorization rules to ensure users can perform only the activities permitted by their roles.
- User acceptance testing (UAT): Confirm that ERP workflows meet business expectations and support the requirements of the teams using the application.
Not every ERP engagement requires the same combination of testing activities. You should prioritize coverage based on application complexity, business impact, change frequency, and identified risks.
How Do You Maintain ERP Quality Through Continuous Testing?
ERP quality requires attention throughout the application lifecycle. Your testing activities should evolve alongside your implementation, integrations, upgrades, and operational changes.
A structured approach helps you maintain visibility and make better release decisions.
- Understand your ERP environment
Start by reviewing your ERP architecture, modules, integrations, customizations, business workflows, and existing QA practices. This helps identify dependencies and critical processes that require focused validation.
- Define your testing strategy around business risks
Identify the workflows that have the greatest operational or financial impact. Establish testing scope, acceptance criteria, dependencies, test environments, and automation opportunities based on your priorities.
- Design relevant test scenarios
Develop test cases that cover standard transactions, business rules, integration points, and exception scenarios. Prepare representative test data to validate different business conditions.
- Execute testing and validate defects
Perform the required functional, integration, regression, migration, and end-to-end testing activities. Document defects, assess their impact, and coordinate with development or implementation teams to validate corrections.
- Report results and support release decisions
Provide stakeholders with visibility into test execution, coverage, defect status, unresolved issues, and outstanding risks. This information helps your teams determine whether an application change is ready for release.
- Maintain regression coverage
As your ERP environment evolves, update relevant test cases to reflect new configurations, workflows, and integrations. Reuse and maintain regression scenarios to support future releases and reduce repetitive manual effort.
Where Can ERP Test Automation Make a Difference?
When your ERP environment undergoes frequent changes, manually repeating the same validation activities can consume considerable QA effort.
Test automation can help you address this challenge by making repeatable testing activities more consistent and efficient.
You can prioritize automation for stable, frequently executed scenarios, such as financial calculations, approval workflows, data validation, and recurring regression tests.
However, automation requires careful planning. Highly customized workflows, frequently changing configurations, and complex business exceptions may require a combination of automated and manual testing.
Your automation strategy should consider application compatibility, test data management, maintenance requirements, and the effort involved in updating test scripts after ERP changes.
The objective is to build sustainable test coverage that supports faster validation without introducing unnecessary maintenance overhead.
Choosing the Right ERP Testing Partner
Your ERP testing requirements may change depending on whether you are implementing a new platform, migrating existing applications, upgrading a cloud ERP environment, or maintaining an established system.
You may need project-based testing support for a specific release or a dedicated QA team to manage ongoing validation.
When evaluating a testing partner, consider their ability to understand your business processes, work across ERP modules, validate integrations, manage regression coverage, and provide clear visibility into testing outcomes.
You should also look for an engagement model that fits your existing QA structure and delivery requirements.
At USM Business Systems, our ERP testing approach focuses on validating application functionality, business workflows, integrations, and data accuracy across your enterprise environment.
Whether you need support for an ERP implementation, upgrade, migration, or ongoing application maintenance, USM – ai app development company, can align testing activities and team composition with your project scope.
As your organization modernizes its infrastructure, you may also need to address the wider application and infrastructure changes associated with cloud adoption. Explore USM’s Cloud Transformation Services to understand how cloud modernization can support your technology roadmap.
Conclusion: Keep Your ERP Environment Reliable as Your Business Evolves
Every ERP change introduces a need to validate what has changed and what those changes may affect. With the right testing strategy, you can identify risks earlier, protect critical workflows, and maintain confidence in your enterprise applications.
If you are planning an ERP implementation, upgrade, migration, or ongoing quality initiative, explore USM’s ERP Testing Services to discuss a testing approach aligned with your business requirements. Let’s Talk.
Frequently Asked Questions
- What are the different types of ERP testing?
The main types of ERP testing include functional testing, integration testing, regression testing, data migration testing, performance testing, security testing, and user acceptance testing (UAT). Organizations may also use end-to-end testing to validate complete business workflows across multiple ERP modules and connected applications.
- How do I choose the right ERP testing services provider?
Look for an ERP testing partner with experience in enterprise applications, complex business workflows, ERP integrations, and data validation. Evaluate their ability to understand your business processes, define risk-based testing strategies, manage test execution, and provide clear reporting on defects and application quality.
- Can ERP testing services be customized for our business requirements?
Yes. ERP testing services can be aligned with your ERP platform, business processes, application architecture, integrations, and project requirements. The testing scope and team composition can be tailored to support specific modules, customized workflows, implementation needs, and ongoing application changes.
- How much do ERP testing services cost?
ERP testing costs depend on the application landscape, number of modules, integration complexity, testing scope, automation requirements, project duration, and team size. Providers may offer project-based, fixed-scope, or dedicated QA engagement models depending on your requirements.
- Can an external ERP testing team work with our existing QA and implementation teams?
Yes. An external ERP testing team can collaborate with your internal QA, development, business, and ERP implementation teams. The engagement can be structured to complement existing resources, provide specialized testing expertise, or manage specific testing activities based on your project needs.
- What deliverables should we expect from an ERP testing services engagement?
Typical deliverables include a testing strategy, test scenarios and cases, test execution reports, defect logs, test coverage reports, data validation results, and release-readiness assessments. The exact deliverables depend on the engagement scope and reporting requirements.
Writers Now Have Easy Access to 400+ AI Models
MuseSpark AI has just rolled-out a service that offers writers and others simultaneous access to more than 400 AI models — including ChatGPT, Gemini and Claude.
That kind of variety is perfect for writers continually cruising the AI universe for Open Source AI that perfectly matches their needs.
Plus, MuseSpark also offers price competition for more established AI model clearinghouses like OpenRouter, which currently features access to 600+ AI models.
In other news and analysis on AI writing:
One Writer’s Take: How to Super-Charge Your Writing Business With AI: Like many writers, Jane Friedman has embraced AI as a tool to enhance her writing business – rather than shrinking from the new tech like a Luddite.
Observes Friedman: “I don’t think prohibiting AI use or shaming people for AI use is productive or necessary for us to live in a world that supports human creativity.”
Click here for a complete rundown on how Friedman gets the most from AI, including:
–Doing better research searches
–Setting up an AI answer chatbot for Web site visitors
–Setting up auto-analysis of site comments
–Setting up auto-analysis of factors driving newsletter subscriptions
–Converting newsletter issues into single articles for a Web site
*Claude AI Will Attend Your Next Writer’s Meeting Live: Writers looking for an AI that will attend a video meeting with them now have Claude Meetings.
Just released, the new feature to Claude is designed to attend a video meeting as an active participant, listen to what is said, answer questions posed to it, keep a record of the meeting for future access — and more.
Claude Meetings currently works with Google Meet and is promised to work with other popular video meeting apps soon.
*Your Favorite New Model for AI Writing Has a Life of Six Weeks: AI is evolving at such a breakneck pace, the heaviest users of AI usually move onto The Next Big Thing in AI models about every six weeks, according to a new study.
Observes writer Christian Cawley: After six weeks, “attention and use of the model decreases as teams and developers proceed to switch to the latest shiny alternative.”
*Turn Your Scripts Into 30-Minute Video Narrations With AI: Bandicam is out with a new text-to-video tool that transforms a script into a multimedia video narration up to 30 minutes long.
Simply paste-in your script and the tool – dubbed Super Narrator — will add a voice-over, visuals, captions and background music to produce a finished video.
Pricing starts at $7.99, which gets you a video that’s nearly 30 minutes long.
*NetSuite Releases AI Auto-Report Generator: Writers already working within the NetSuite ecosystem will want to check out new AI that auto-generates reports from data stored within NetSuite.
The new tool is designed to quickly put together reports triggered by natural conversation – without the need for you to understand how to integrate saved searches, formulas or report configurations.
Essentially: Users simply describe the report they need and the tool’s AI asks follow-up questions to build the desired report.
*Firmly Entrenched: Once Unimaginable, AI-Written Books Are Becoming the Norm: Self-published writers are apparently leading the charge when it comes to using AI to churn-out fiction and non-fiction, according to a new study from Stony Brook University.
These days, about a fifth of all self-published books feature text that’s at least 25% generated by AI, according to the study.
Observes writer Sophie McBain: “One literary agent told me that when it comes to AI, the atmosphere in mainstream publishing is like ‘the wild west.’ The technology is outpacing the industry’s capacity to reckon with what it means that authors can – and already are – using AI to write their books.”
*Automated Email Writing: Too Much of a Good Thing?: Writer Renee Giarusso reports that so many people are using AI to write long and in-depth emails these days, recipients are suffering from content overload.
Observes Giarusso: “As I always share with leaders, create your own IP first, then use AI to curate it, condense it, fact find or research — not to multiply the noise.”
*Writers Lose a Key Champion of AI Creative Writing: Nvidia’s decision to oust its CEO Jensen Huang means the Open Source AI community has lost one of its strongest advocates.
As the founding force behind Nvidia – which has a near monopoly on the chips powering the AI revolution – Huang was also a strong advocate of Open Source AI, which currently offers some of the most creative AI writing options on the planet right now.
Observes writer Rebecca Ballan: “Huang has been a staunch supporter of Open Source AI, framing the ecosystem as good for safety, innovation, diffusion and sovereignty.”
*AI Growth May Slow With Ouster of Nvidia CEO Jensen Huang: One of the unintended consequences of the loss of Jensen Huang as CEO of powerhouse AI chipmaker Nvidia may be slower overall growth of AI.
The reason: For years, Nvidia under Huang had been known as the ‘central bank of AI,’ loaning billions of dollars to scores of corporate AI users and AI startups in an effort to repeatedly supercharge the tech’s growth.
That’s expected to change under the new Interim CEO of Nvidia, Colette Kress – formerly Nvidia’s chief financial officer — who has expressed discomfort at Huang’s heavy handed financing style.

Share a Link: Please consider sharing a link to https://RobotWritersAI.com from your blog, social media post, publication or emails. More links leading to RobotWritersAI.com helps everyone interested in AI-generated writing.
–Joe Dysart is editor of RobotWritersAI.com and a tech journalist with 20+ years experience. His work has appeared in 150+ publications, including The New York Times and the Financial Times of London.
The post Writers Now Have Easy Access to 400+ AI Models appeared first on Robot Writers AI.
AI discovers a hidden heart signal that could predict sudden death
How Artificial Intelligence Is Reshaping the Legal Profession
Artificial intelligence is no longer a concept confined to science fiction or Silicon Valley boardrooms. It has quietly worked its way into courtrooms, law offices, and legal research departments across the country. For attorneys, paralegals, and clients alike, the rise […]
The post How Artificial Intelligence Is Reshaping the Legal Profession appeared first on Techspective: A Unique Perspective on Technology.
Robots recognize the shape of an arena by messaging each other
AI supercharges laser physics with a stunning 250x speed boost
Robot Talk Episode 170 – Robots in the real world, with Daniela Shchegelskaya
Claire chatted to Daniela Shchegelskaya from the University of Twente about getting robots to operate reliably in safety-critical environments.
Daniela Shchegelskaya is an electronics and robotics engineer specialising in autonomous systems, control engineering, and Al-driven robotics for industrial and safety-critical environments. She works on deploying real-world autonomous robots beyond lab conditions, focusing on reliability, connectivity, and system-level design. At Elimpus, she works on the integration of Boston Dynamics’ Spot for inspections in live power substations, including autonomous navigation, network connectivity, remote diagnostics, data collection and system-level reliability.

