Intelligent Robotics — Investing in Minds Not Bots
The Timing is Right
Conventional industrial robots do one or two pre-programmed jobs very well. They’re reliable, limited, expensive and come with complicated user interfaces. Intelligent robots couldn’t be more different. Thanks to a perfect storm of certain technologies converging (namely: deep learning, neural networks, sensors, parallel processors, computer power, advanced computer vision and inexpensive hardware) the age of robotics is becoming a reality.
There have also been clear advances with personal or consumer robots. Whilst these are entertaining, educational and even a tool to combat loneliness (for example: Robo Wunderkind; Bots & Us; MyVan Robotics; No Isolation), they are further away from mass adoption than intelligent robots aimed at the enterprise market.
Frost & Sullivan anticipate that 35% of leading organisations across all sectors are likely to adopt intelligent robots in their daily operations over the next five years. BCC research forecasts demand for robots and their associated products will increase at a compound annual growth rate of four percent until 2021, reaching $31.5 billion. The most rapid integration into workplaces is forecast for the European market, which BCC forecasts will see a compound annual growth rate of six percent (above the four percent global average), although Asia-Pacific will end up representing $12 billion of the $31.5 billion total market (c.38%). Other research puts this figure far higher.
What about China?
“Made in China 2025” outlined China’s ambitious plan to turn the country into a manufacturing superpower targeting all high-tech industries including, not surprisingly, robotics. This paved the way for the country’s 13th five-year plan which outlined a key objective being “innovative-led growth”. Since then, we have seen China aggressively driving the adoption and manufacturing of robots. R&D activity has soared and there has been a 95% increase in patent activity in intelligent robotics in China since 2016. The size of the domestic market, pressure on companies to automate and support of the government through financial incentives has transformed the sector.
But we still believe there is space for the UK and Europe to produce its own valuable businesses in this sector.
The UK Market
The UK is currently behind other advanced nations in overall productivity (output per worker). This is partly due to lower level of adoption of digital and automation technology and is particularly acute amongst SMEs. Services industries (distribution, transport, communication, finance and business services) contribute c.79% to UK GDP; production (mining, manufacturing, waste management and water supply) 14%; construction 6% and agriculture 0.7%. It is clear that there are opportunities in each of these industries and the UK Government is looking to raise the country’s productivity and international competitiveness through, amongst other initiatives, R&D leadership, trusted data exchanges and targeted financial incentives to promote investment in this area. The Government has also recently announced $1.4 billion investment in artificial intelligent technologies.
Marry the support and buy-in of the UK government and its European counterparts (like SPARC, under the Horizon 2020 framework) with the unparalleled talent in European universities, such as Oxford, Cambridge, Imperial, UCL, KCL, Edinburgh, ETH Zurich, Manchester, Munich (TUM) and Bristol, and we expect there to be many exciting opportunities to invest in this space.
There are multiple market drivers propelling adoption of intelligent robotics across enterprises. Firstly, wherever you are in the world, China included, it is becoming more and more difficult to recruit and keep skilled workers. In looking to plug this gap, modern robots are attractive. They are able to do physically demanding work in challenging environmental conditions with increasing reliability, precision and quality, at a lower operational cost, with improved productivity. The functionalities offered by new robots are advancing, whilst the costs are falling. Robots aimed at SMEs are becoming affordable. The ability for organisations to redeploy human workers into more creative activities, ultimately adding greater value, can result in tangible business benefits.
Despite the opportunities, this space is tough for startups. Challenges remain in the following areas:
- Customer adoption: With pure software, customers now largely understand how their businesses can benefit. Robots are more complicated. They require, even at a basic level, a physical structure, hardware to sense and interact with their surroundings, software to interpret the inputs being collected and signals to control their motors and actuators. Customer fears centre around:
- High initial investment costs: set-up and integration with existing infrastructure.
- Requirements to upgrade: sensors, circuits, software, hardware components that can stop performing after certain usage time.
- Ability to train employees to use or work alongside the robots as well as perennial PR worries around labour replacement.
- Security and privacy: robots will be connected to distributed control systems with the potential for intruder influence.
- Performance and reliability: intelligent robots are often intended to work in a wide range of conditions. They must perform a broad range of functions, which makes them difficult to test. It isn’t always clear that performance obtained under one set of conditions will translate to another situation. And because the inherent technologies are so new, customers are concerned about how long the hardware will take to fail, particularly in challenging real-world environments.
- Health and Safety: where robots co-exist with human workers, design must account for ISO standards, CE certification and basic safety.
Building on the above, it is therefore crucial for a startup to be able to demonstrate real business value and not be tempted to chase the “wow factor” (although watching Atlas somersault is, undeniably, very very cool).
Entrepreneurs also face obstacles:
- Enabling technologies such as advanced reasoning, perception, control and manipulation, may still not be as advanced as required for a robot to undertake a broad set of activities that certain teams are trying to attain.
- Creating prototypes and conducting trials and pilots to collect data, prove ROI and iterate design on the basis of customer feedback can take longer than originally anticipated.
- In contrast to software design, hardware design is much more expensive in terms of time-cost and technology risk. Costs only increase as a design matures.
- There is a disproportionately large cost and inherent difficulty associated with producing robotic hardware in small volumes. There is a chicken and egg problem here: you can’t achieve scale without being cost effective but you cannot be cost effective without scaling. We often see the projected BOM (bill of materials) drop significantly as 10s and 100s of robots are produced.
- Scaling up manufacture and assembly raises questions: will you outsource to a contract manufacturer? How will you ensure quality control at that point? Will that be in Asia, and if so, how is your local knowledge out there?
- Finding funders who can see through the journey beyond seed and series A remains difficult. Whilst capital in the VC and growth markets is increasing, large chunks of the market would simply not look at startups in this area. They are not familiar with the business model and see the potential capital required as unappealing.
All of this sits alongside the usual entrepreneur worries of building and retaining excellent teams. In this field, they must also be technologically capable. Then there’s the issue of putting together a sensible go-to-market strategy and executing against your plan without running out of cash. Certainly not straightforward.
Despite these challenges, we have already seen some success stories in Europe. For example: Teradyne acquiring Universal Robots in 2015 and then Mobile Industrial Robots in 2018; Midea’s takeover of Kuka in 2016 and Cambridge Medical Robotics receiving equity investment of over $100m as at June 2018.
Exciting European startups
We have seen startups exploring the following areas:
- Software: Higher levels of autonomous behaviour require situational awareness of the internal state of a system and its environment. For example, the ability to plan a series of tasks and, if mobile, map a path on which it travels. At a very high level of sophistication, a robot may be able to develop its own strategies and rules to achieve a particular output. We have seen really interesting software solutions for autonomous vehicles, taking sensor inputs and understanding localisation, perception and planning and outputting control instructions. In the same vein, we have also seen software products for robots and drones which take information from sensors to accurately position themselves within a 3D space, accurately map the space around them and have a semantic understanding of that space. [SLAMcore; Oxbotica; Kudan]
- Cobots: Lighter mobile robots able to perform tasks in and around humans, particularly repetitive tasks that are part of the manufacturing and assembly supply chain. We have seen cobots based on the Universal Robot arms but with a consumer-friendly interface, these are easier to install at a more affordable price point. [Automata Technologies; Franka Emika]
- Warehouse and Logistics: With the surge of e-commerce and on-demand delivery, we have seen robots in many guises that can assist in a warehouse environment. Whether this is an autonomous mobile robot which picks up and delivers pallets, to one which specialises in picking and placing, or rapidly sorting items from a production line, there are multiple ways in which warehouses can become more efficient and dynamic. [Gideon Brothers; NoMagic.ai]
- Labour scarcity: Acting on a pull from the end-customers, we have seen startups tackling the difficult task of putting autonomous robots into relatively unstructured environments such as fields, polytunnels and glasshouses, where they can identify, inspect, pick and place fruit. We have also seen autonomous robots moving freely within fields, detecting and precisely spraying herbicide. [Xihelm; Dogtooth; Ecorobotix]. We have seen robots which can enter small spaces to carry out inspection functions in houses. These can be applied, for the example, in the area of insulation. There is also creativity around the movement of the robot and its ability to sense effectively in dusty, dirty or otherwise hostile environments. [Q Bot; Zoa Robotics]. Finally, we have seen robots intending to be utilised in the service industry eventually aiming to become part of the fast food and drink workforce [Barsys; Dex.ai].
What we are looking out for
Given our interest and enthusiasm, we would certainly love to meet any startups exploring intelligent robotics. In this sector, we will look for:
- Team: a founding team that combines deep technical expertise and an understanding of building and scaling their products with commercial expertise. Identifying your target market and early adopter customers with a clear route into them is critical.
- Early partnership with customers: as above, we expect that your design, KPIs and approach will be significantly informed by your proof of concept customers. Identifying a heavy-weight in the industry who is open to testing your product and giving you feedback will be useful.
- Empathetic design: we expect you to have considered the practical uses of your product. You will have accounted for the fact that operators of your robot may not be engineers. Intuitive, easy-to-use interfaces are the goal. We do not expect other systems to have to be retrofitted to accommodate your robot.
Exciting changes on the horizon (short and longer-term):
- Voice, gesture, tactile and general visual inputs to form part of the UI with multimodal input/output variations.
- Robots inspired by nature (including humans!) involved in last mile logistics deliveries and inspection, or acting together collaboratively in swarms.
- Advances in motors and sensors resulting in greater ranges of motion, less battery re-charges and change-outs and advanced manipulation through haptic sensors.
- Soft and light materials being used in design along with materials which harvest and store energy.
- Robots which can be used in challenging and harsh environments (note Headlight), including deep sea and space.
We’d love to hear from you — please do email me [email protected].
Thanks to Duncan Robertson for reading a draft of this blog and to the following folks who have had conversations with me and/or let me explore their fantastic companies — @mostafa_eSayed; @automata_tech; @SteveSchlenker; @sethwinterroth; @dexairobotics; @kacpernowicki; @gideonbros; @matijakopic; @milanracic; @iendeavors; @accdynamics; @itsjasonblack; @NpappaG; @openbionics; @joelgibbard; @KudanSLAM; @SLAMcoreLtd; @oxbotica; @frankaemika; @ecorobotix; @join_ef; @zoa_robotics; @mhoward3210; @sparc_eu; @eu_h2020; @innovateuk.