#96 IP and Robotics & Autonomous Systems

00:00:02: Welcome to the IP Market Signal series from IP Management Voice.

00:00:06: Based on OFB, IP market research group analysis this series explores the industries technologies and market developments that are shaping the demand for intellectual property expertise.

00:00:19: each episode highlights key topics driving current discussions influential market voices emerging needs among innovators in businesses an opportunities for ip experts created by these trends.

00:00:32: In this episode, we explore how IP market signals are shaping the fields of robotics drones and autonomous systems.

00:00:41: You know usually when you talk about software.

00:00:43: We think of it as something that is Trapped behind a screen like your type-of command if processes some data Maybe a spits out a spreadsheet or generates an image right?

00:00:53: It's

00:00:53: completely confined to the digital realm exactly just lives on a server somewhere totally abstracted from the messy reality of gravity, friction or unpredictable human beings.

00:01:05: But looking at the stack of research and market intelligence we've gathered for this deep dive into the European intellectual property landscape here in the summer of twenty-twenty six software has grown legs

00:01:16: yeah and wheels

00:01:17: and robotic arms.

00:01:19: We are exploring the era of embodied intelligence Software that doesn't just calculate but actually senses its environment makes a real time decision and acts in the physical world.

00:01:29: Which is a massive shift!

00:01:30: It IS, And for you, The Intellectual Property Expert listening today.

00:01:33: this evolution is basically rewriting every single rule of your playbook.

00:01:48: Right?

00:01:48: Like patenting a novel hinge that's largely behind us.

00:01:52: The commercially decisive value doesn't sit in single isolated mechanical component anymore.

00:01:58: So where is it sitting?

00:01:59: It sits entirely in the interactions.

00:02:01: We're looking at the interaction between a LiDAR sensor and a hydraulic actuator or, uh...the real-time negotiation between an AI model and hardcoded safety rules.

00:02:12: Wow!

00:02:12: Or even the communication protocols between single autonomous robot and its entire fleet.

00:02:17: For IP professional this requires fundamental shift from traditional component protection to strategic system control.

00:02:24: Okay Let's unpack this, because that is a massive pivot for the industry.

00:02:27: To really understand the mechanics of where this value is shifting we should probably look at a sector that has had a long time to mature.

00:02:34: That makes sense.

00:02:35: Take surgical robotics I mean.

00:02:37: For years The narrative was essentially A mechanical arms race.

00:02:41: You know who could build the most precise Least invasive Mechanical scalpel?

00:02:45: Yeah but if you Look At landscape now...That battleground Has completely moved.

00:02:50: Moved Where?

00:02:51: It has migrated almost entirely to the digital and ecosystem layers.

00:02:56: Actually, let me bring in a brilliant observation on this from Jack Severs at Gil Jennings and everything.

00:03:01: What does he say?

00:03:02: He points out —and I'll quote him here—

00:03:11: That is a really fascinating metric.

00:03:13: Yeah, so the innovation in the actual physical joints.

00:03:16: you know The motors that grip strength?

00:03:17: That's plateauing.

00:03:18: maturing Is probably the better word for it.

00:03:20: the kinematic equations and the core mechanical engineering challenges of moving a robotic arm smoothly.

00:03:27: They've largely been solved.

00:03:28: So hardwares basically good enough.

00:03:30: exactly sever sees the surgical robotics value shifting away from core hardware And moving toward navigation simulation training and workflow integration.

00:03:39: okay that makes sense.

00:03:40: Think about what a surgeon actually needs today.

00:03:43: They don't need a sharper scalpel, they need an AI system that can take a patient's preoperative MRI overlay it in real time onto alive three D video feed of the surgery and use haptic feedback to physically resist the surgeons hand if they get too close to a major artery.

00:04:01: Oh wow!

00:04:01: Yeah That software driven navigation and training simulators required to teach surgeons how do you use it?

00:04:10: differentiable value sets now.

00:04:12: That reframes the entire product.

00:04:13: I mean you aren't selling a robot anymore, You are selling an augmented surgical environment

00:04:18: precisely.

00:04:19: and that actually brings up a secondary challenge which Pamela Breyer from Marks & Clerk framed perfectly when she was analyzing the UK's health care system.

00:04:26: oh

00:04:26: right the ecosystem

00:04:27: aspect yeah She stated investment in robotic surgical systems has the potential to invigorate the NHS.

00:04:34: However it is equally important to consider the infrastructure including clinician training.

00:04:39: It's so true.

00:04:40: Briar frames successful surgical robotics adoption as an ecosystem challenge.

00:04:45: it requires protection and investment not just in the devices but in training, all that surrounding infrastructure.

00:04:52: Because if a hospital buys a ten million euro surgical robot But data integration is poor or the training modules aren't optimized The machine sits idle

00:05:02: Exactly!

00:05:03: The IP strategy has to cover proprietary training software the user interface, and data pipelines just as aggressively as robotic chassis itself.

00:05:12: I see logic there but let me challenge this for a second... If hardware innovation is leveling off does that mean physical machines themselves are becoming generic commodities?

00:05:22: Right to big fear!

00:05:23: Yeah

00:05:23: if i am a hardware manufacturer might just destined to sell dumb metal while software companies capture all high margin value.

00:05:30: Well that's existential fear keeping hardware executives awake at night.

00:05:35: But the reality of physics prevents hardware from becoming completely commoditized, especially in extreme environments.

00:05:41: Extreme environments like space?

00:05:43: Space sure but let's look at subsea robotics.

00:05:47: The ocean is incredibly hostile to technology.

00:05:50: Radio waves do not travel well through saltwater which means you cannot rely on cloud computing.

00:05:56: Of course A subsea robot can't phone home a massive server farm process image data.

00:06:02: The latency of acoustic modems is just far too slow.

00:06:05: So the intelligence has to live entirely on board this submarine?

00:06:08: Precisely, and that's where Darina Slavova from Muber & Ellis provides a really crucial perspective.

00:06:14: She highlights company called Frontier Robotics To illustrate how this constraint drives innovation.

00:06:19: Okay what was her assessment there?

00:06:20: She says quote.

00:06:21: This software-driven approach positions Frontier robotics at the forefront Of shift towards scalable cost effective subsea automation.

00:06:29: Break down the mechanics of that software-driven approach for us?

00:06:32: How does a software focus solve physical limitations in the ocean?

00:06:36: Slovova presents software retrofits, edge computing and real time mapping as scalable routes to autonomous subsea inspection.

00:06:43: And crucially they do this without replacing existing vehicle platforms.

00:06:48: Edge computing...

00:06:49: Meaning data is processed at the end right there on the robot's local hardware, rather than in some distant cloud server.

00:06:56: Instead of designing a completely new incredibly expensive underwater vehicle from scratch they take conventional remotely operated under water vehicles and upgrade them into fully autonomous systems using digital retrofits.

00:07:10: It is!

00:07:11: The Hardware isn't generic commodity here.

00:07:14: it has highly specialized pressure resistant vessel that absolutely essential for housing this advanced localized.

00:07:22: What's fascinating to me here is how this shift To software and ecosystem value.

00:07:27: it triggers massive really complex legal battles When the value moves away from a single physical part And just spreads across the entire software stack competitors inevitably step on each other's toes.

00:07:38: Oh

00:07:38: constantly.

00:07:39: what's fascinating?

00:07:40: Here, Is how The recent warehouse robotics patent war between okado an auto store.

00:07:45: It's Just A prime example of This.

00:07:47: yes

00:07:47: that dispute was a masterclass in modern IP warfare.

00:07:50: They were fighting over automated storage and retrieval technology.

00:07:54: Essentially, the software logic... ...and physical grid designs that allow hundreds of robots to navigate a warehouse simultaneously without

00:08:01: colliding.".

00:08:02: And they fought everywhere!

00:08:03: Across U.S.

00:08:04: International Trade Commission, UK High Court German Courts European Patent Office.

00:08:09: But resolution is what really matters for our listener.

00:08:12: I think Right

00:08:12: because they settled globally

00:08:14: Exactly.

00:08:15: Why?

00:08:15: Because they realized that cross-licensing their pre- Twenty-Twenty patent portfolios was the absolute only way to ensure both companies could actually keep building products.

00:08:25: They needed freedom to operate across the entire robotic stack more than they needed some Pyrrhic victory on one specific tiny component.

00:08:33: Which brings us to the core dilemma for you, The IP Expert If software control loops and AI are actual drivers of commercial value how do you legally protect them?

00:08:43: It's the million-dollar question.

00:08:45: European patent law famously does not allow you to patent an abstract algorithm or a mathematical method on its own.

00:08:51: Right, You can't just walk into The European Patent Office with A brilliant line of code On a flash drive and Walk out With a patent.

00:08:58: Yeah There has To be a physical Manifestation Of that Codes Utility.

00:09:02: Muhammad Amadi Bidak Vidi From VO patents & Trademarks Actually Sets The legal Baseline For This Incredibly Well.

00:09:08: What Is His Rule?

00:09:09: He says, an AI invention is patentable if it explicitly realizes a technical use or improvement.

00:09:16: A technical user improvement?

00:09:18: Let's define what the patent office actually considers technical in this context.

00:09:22: Bidak Fiti treats real-world technical integration and a demonstrable technical effect as the decisive route to patentability for AI enabled autonomous

00:09:31: systems.".

00:09:32: So an example would be...

00:09:33: If your AI model is just processing financial data to predict currency fluctuations, the Patent Office views that as an abstract business method.

00:09:41: It's not patentable!

00:09:43: However if you're AI model interpreting LIDAR data to optimize the torque of a robotic arm thereby reducing motor energy consumption by say fifteen percent That energy reduction is demonstrable real world technical effect

00:09:57: Here where it gets really interesting though Because a robot operating in the physical world has severe limitations that a cloud server just doesn't have.

00:10:06: We touched on this with the subsea example, but it really applies everywhere.

00:10:11: Absolutely every

00:10:12: one drone as a tiny battery an agricultural robot bakes-in-the-hot Sun and Has a very limited computing space onboard

00:10:19: which is exactly why?

00:10:21: The physical limitations of the hardware actually become the secret weapon for patenting the software.

00:10:26: oh I like that framing.

00:10:27: Fabian Kindle from Maywald articulates this perfectly.

00:10:31: He states,

00:10:43: Exactly!

00:10:44: Kindle argues that AI patentability in robotics and autonomous systems actually depends on protecting the concrete engineering adaptations.

00:10:57: Exactly, think about a massive neural network trained in the state-of-the art data center.

00:11:02: To get that model to actually run on small low power chip inside of drone engineers have to

00:11:08: compress it.

00:11:09: They use techniques like quantization which is reducing mathematical precision so requires less memory or pruning unnecessary neural connections.

00:11:19: Okay, so you don't try to patent the abstract AI model itself?

00:11:23: No no!

00:11:23: You patent this specific quantization technique... ...you patent thermal management software or memory allocation protocols that allow that AI function on a ten-watt chip without melting the drone.

00:11:34: Let's take this concept out into the wild.

00:11:36: Agrotech.

00:11:37: Agricultural robotics.

00:11:39: You have autonomous tractors and computer vision.

00:11:42: weed picking robots operating in muddy totally unpredictable fields.

00:11:47: When involvement is entirely chaotic, that

00:11:49: chaos forces a highly strategic division of your IP portfolio doesn't

00:11:53: it?

00:11:53: It does.

00:11:54: Ben Hunter from D Young & Co sums up the agri-tech challenge beautifully he says.

00:11:58: in practice That contribution often arises from The way the AI model is applied to A real world technical problem.

00:12:04: so the Contribution isn't the code itself.

00:12:07: its how the AI Navigates the mud the uneven terrain the changing light conditions Of the Sun as it moves across the sky.

00:12:14: exactly.

00:12:15: Hunter argues that Agrotech AI patents must be anchored in a specific technical effect while the commercially sensitive training data and the coal model architecture are protected selectively as trade secrets.

00:12:28: Let me make sure I'm visualizing this correctly, we aren't trying to patent the master recipe for the AI itself because publishing that pattern would give our competitors exact instructions to recreate our intelligence.

00:12:41: Right, you just hand them the playbook!

00:12:42: Instead we are patenting that highly specific energy efficient oven we built...to bake that exact recipe in the middle of a muddy field.

00:12:50: The Muddy Oven Analogy is absolutely spot on!

00:12:57: What?

00:12:58: That is the recipe, and you lock it in a vault as a trade secret.

00:13:02: But the specific way your software regulates the tractor's hydraulics based on AI output.

00:13:07: that's the oven!

00:13:08: And this goes into patent specification.

00:13:12: If I rely on trade secrets for my core AI model, couldn't a competitor just buy one of my autonomous tractors put it in their own field monitor how it behaves and reverse engineer My AI logic?

00:13:25: Just by watching it.

00:13:26: Well

00:13:27: that is the inherent risk Of trade.

00:13:28: secret sure but reverse engineering A complex neural network simply By observing its physical outputs Is nearly impossible.

00:13:36: really even with advanced tools.

00:13:38: a Neural Network is a black box Even to its creators.

00:13:41: sometimes A competitor might see that your tractor avoids a rock, but they cannot see.

00:13:46: the millions of weighted mathematical connections led to this specific decision.

00:13:51: The real risk isn't a competitor watching the tractor in field – the real risk is an employee walking out door with USB drive containing training data.

00:14:01: Exactly!

00:14:02: That's why robust internal data governance is just as critical for IP experts today.

00:14:08: That transitions us perfectly into our next major challenge.

00:14:11: Yeah, because we have spent this entire time talking about the IP strategy for a single machine one tractor One submarine yeah But the complexity multiplies exponentially when these machines stop acting alone and start operating in swarms Particularly in defense and security applications.

00:14:29: The shift from single agents to swarm robotics is arguably the most significant technical leap we are seeing today.

00:14:36: And

00:14:36: the capital flowing into this space totally reflects that, I mean quantum systems in Munich recently raised one point two billion dollars.

00:14:43: and when you look at the macro trends in patent filings for drone swarms Chinese universities and institutions are just dominating the landscape.

00:14:51: The patent data is a glaring indicator of geopolitical and market priorities right now.

00:14:56: Andrew White from Mathis & Squire tracks this evolution closely.

00:14:59: What does he say about

00:15:00: it?

00:15:00: He notes, and I quote, It

00:15:09: is undeniable when you look at how autonomous aerial systems are being deployed—I mean they're monitoring critical infrastructure or operating in conflict zones but mechanically.

00:15:18: what makes a swarm so different?

00:15:22: Ten drones at once.

00:15:24: It comes down to a concept called multi-agent reinforcement learning combined with heterogeneous swarm task allocation.

00:15:31: Okay, we definitely need to translate those two terms for the audience.

00:15:34: Let's break it down.

00:15:36: If you fly ten individual drones A human operator is still giving ten separate commands.

00:15:41: Right But in a true swarm You give one high level command To the group Like secure this perimeter

00:15:47: And they figure out.

00:15:47: how do I?

00:15:48: Yes

00:15:49: Multi-agent reinforcement learning allows the drones to communicate locally with each other, acting like a flock of starlings.

00:15:54: They negotiate in real time to achieve their goal without central command

00:15:59: and the heterogeneous task allocation.

00:16:01: That means that the drones have different jobs.

00:16:03: One drone carries radar jammers Another carries thermal cameras And third acts as communication relay And they autonomously decide which drone does what job based on changing environment

00:16:16: wild.

00:16:17: So White Reads rising drone and counter-drone filings as evidence that autonomous aerial systems have become a mainstream security in infrastructure protection market?

00:16:27: Absolutely!

00:16:28: As an IP professional, you aren't just patenting a drone—you're patenting the decentralized communication protocols.

00:16:38: Now if we connect this to the bigger picture, This level of autonomy brings us directly into collision with European regulators.

00:16:44: Oh boy!

00:16:45: Because an autonomous swarm or even just a single self-learning robot is a regulatory nightmare.

00:16:51: Which brings us to EU machinery regulation which becomes fully applicable in January twenty-twenty seven.

00:16:57: This piece legislation fundamentally alters how hardware and software interact under law doesn't it?

00:17:02: It really does.

00:17:03: Historically safety regulations applied to machine as left factory floor.

00:17:07: The manufacturer approved it was safe, slapped a CE mark on and sold.

00:17:10: Done!

00:17:12: But the new machinery regulation brings autonomous mobile machinery and self-evolving AI into the scope of CE marking.

00:17:18: Explain the mechanics to this self evolving aspect.

00:17:21: Let's say your robotics company deploys fleet warehouse robots.

00:17:26: Robots learn from navigating around human workers for...a month.

00:17:30: The company takes that data improves the AI model And pushes an over the air software update to make them more efficient.

00:17:39: What happens

00:17:41: then?

00:17:43: The company pushing the update basically becomes the legal manufacturer of a newly modified machine.

00:18:07: That

00:18:07: is staggering!

00:18:08: Pushing a line of code requires physical safety recertification, and this introduces a concept we see heavily emphasized in research...the data learning loop.

00:18:17: Yes…the

00:18:17: loop is everything.

00:18:19: Real-world operation generates data.

00:18:21: That data is fed back to improve model – improved models deployed for updating system.

00:18:27: For an IP professional Why does this safety compliance loop matter so much?

00:18:31: Because the technical documentation that the engineering teams must build to satisfy the machinery regulation and the overlapping EU AI Act, well.

00:18:40: That is the exact same documentation that serves as your IP relevant disclosure evidence.

00:18:46: You cannot have a compliance team telling regulators one thing about how they make safety decisions while The legal team tells the patent office something completely contradictory.

00:18:55: Wow

00:18:56: This means the concept of freedom to operate, FTO is just completely revolutionized.

00:19:00: I mean FTO used to be a discrete task.

00:19:03: before launching a product you search existing patents To ensure your new robot doesn't infringe on anyone else's IP.

00:19:09: check The box and move on.

00:19:11: that static approach to fto Is dead because

00:19:13: the robot is no longer static if Your system is constantly ingesting new open-source code libraries for its vision system executing over-the-air updates that alter its behavior and changing how it communicates with third party infrastructure.

00:19:28: Your freedom to operate status is shifting on a daily

00:19:32: basis.".

00:19:33: It is, this raises an important question about systemic governance system.

00:19:37: FTO must be a continuous living process.

00:19:40: How do you even manage?

00:19:41: That

00:19:42: has to map the proprietary hardware The AI model weights the open source software licenses the communication protocols And the real world deployment parameters simultaneously.

00:19:52: If an engineer downloads a new open-source module to improve the drone's camera resolution, that single action might compromise the entire system's FTO.

00:20:01: So what does this all mean?

00:20:02: if we synthesize this stack of research for the IEP professional listening right now The main takeaway is your job description has undergone radical transformation.

00:20:10: Radical is the right word.

00:20:11: You are no longer just legal mechanism for securing patent claims on finished product.

00:20:16: you're strategic architect.

00:20:18: Yes, you are creating decision capability for management.

00:20:23: You have to look at a highly complex embodied intelligence system and map out the entire stack.

00:20:29: You decide what the company must explicitly disclose to the patent office... ...to secure a monopoly on physical application.

00:20:35: Exactly!

00:20:36: You determine what core data architectures must be locked down as trade secrets….

00:20:41: …you figure out what communication protocols should be openly licensed to encourage third-party ecosystem growth.

00:20:47: You are the architect maintaining strategic control across mechanics, software and data.

00:20:53: It requires a fluency that bridges mechanical engineering, data science and regulatory compliance.

00:20:58: But before we wrap up there is a geopolitical consequence to this shift that we need to address.

00:21:03: A

00:21:03: geopolitical consequences?

00:21:04: Yes We discussed how agricultural drone swarms use multi-agent reinforcement learning to inspect crops Right!

00:21:10: but fundamentally the software logic required to coordinate drones over cornfield Is remarkably similar to the logic required To coordinate a drone swarm in military conflict zone.

00:21:20: The technology is inherently dual use

00:21:23: Exactly.

00:21:24: And this leaves you with a final thought to mull over.

00:21:27: as an IP professional, You might build the perfect portfolio patenting The hardware constraints protecting the AI training data As a trade secret and maintaining continuous FTO

00:21:38: Doing everything right

00:21:39: doing Everything Right.

00:21:40: but if the embodied intelligence you protect becomes so advanced that it is gained critical national security asset What happens to your intellectual property rights?

00:21:49: Oh wow We are entering an era where governments could invoke national security laws to enforce compulsory licensing, or outright seize the patents governing these autonomous swarms.

00:22:01: What if The Robot of Tomorrow is essentially owned by state because end users inadvertently trained it as a security asset?

00:22:08: That's a chilling scenario!

00:22:10: You can successfully navigate entire business and regulatory landscape only to have the underlying technology seized, because The Sovereign State realizes your robotic ecosystem is too powerful to remain purely in private hands.

00:22:22: A

00:22:22: real possibility!

00:22:24: When the boundary between a commercial tool and a national security asset evaporates... ...the true strength of an IP portfolio will be tested in ways we've never seen.

00:22:33: Thank you for joining us on this deep dive.

00:22:35: Keep questioning the layers of control In the autonomous systems around You And We'll catch you next time.

00:22:42: This audio briefing is based on a comprehensive IP market report produced by the OFB-IP Market Research Group.

00:22:52: If you'd like to explore this topic in greater depth, You can download full reports using link provided in show notes.