The Technology (in more detail)

PeykBot is intelligently designed considering every parameter required for an efficient and sustainable delivery. It is built on a 6 wheel drive all terrain chassis that can sidereal kerbs, 2 climb up to 2 stairs and drive in off-road environments. Its durable composite body and aluminium chassis gives the advantage in offering the highest payload in any delivery robot currently in the market.

To prevent operational restrictions, PeykBot is fundamentally designed as a modular robot that allows Peyk and other third party manufacturers to connect different customised attachments on to the robot. This creates diversity allowing Peyk to expand into different markets and use cases as well as reduces the potential risks of restricting the product into just a single industry.

The base of each peykBot looks like this. Its where all the exciting technology is stored. You will have any of the below attachments connected to this base of the robot.

We did not restrict the peykBot to only fulfill deliveries With this attachment, the robot can be used for indoor and outdoor advertising purposes to display custom content whilst driving. This is exclusively designed for conferences and exhibitions.

PeykBot is primarily used to deliver parcels within an outdoor environment. Due to its autonomous nature, a significant percentage of the robots costs will be dedicated to the implementation of self driving capabilities. Research shows that on average, 55% of the production cost of an autonomous robot is due to the price of the sensors and hardware used (Agnetwork, 2022). Hence, Peyk has identified this as a problem that needs to be addressed.

Our solution is to design a self-driving autonomous platform that reduces its reliance on expensive hardware. To implement this, several algorithms will be used on a software basis to fuse data from lower end sensors in order to process an accurate navigation.

The solution is to eliminate the need of expensive hardware such as 3D Lidar and the prerequisite of pre-mapping in autonomous mode. To create an accurate and cost effective navigation, we will fuse data from a stereo camera, an affordable 360 degrees range scanner, GPS and some ultrasonic sensors. By using this data, the algorithm can generate waypoints from raw RGB-D images and make the optimal actions to reach a goal area while avoiding static and dynamic obstacles. A Deep Reinforcement Learning (DRL)-based method will be used that enables a robot to execute navigation tasks in multi-obstacle environments with randomness and dynamics. Deep reinforcement learning paradigms show great promise for developing navigation stacks in uncertain stochastic environments. They help resolve issues like sensor noise and generate optimal steering policies. Even though many simple algorithms like SLAM or point-cloud are used for automated navigation, they are proven to be computationally much heavier than deep RL. However, this method achieves a state-of-the-art success rate, with at least a 30% improvement of average episode collision through the use of Python and Tensorflow for creating and running the algorithm.

Above is a full scale diagram of the entire journey from data to travel execution.

Above images are examples of how the sidewalks, crosswalks and obstacles are detected by the local robot algorithm.

In the UK, the regulatory environment for autonomous delivery robots is complex and underdeveloped, lacking clear and harmonized rules. Delivery robots are subject to a variety of local, municipal, and national regulations that indirectly address their operation on sidewalks and public spaces. This has led to a situation where deployments are often confined to specific areas under close supervision by authorities like the Department for Transport. The UK's approach has been described as "soft law," relying on flexible guidelines and negotiations between companies and regulatory bodies to allow for pilot schemes. However, this has resulted in a patchwork of rules that vary significantly across jurisdictions, reflecting a broader challenge of integrating autonomous delivery robots into public spaces without comprehensive legal frameworks.

The EU faces a similar challenge, with a lack of uniform regulations for autonomous delivery robots across member states. Only Austria explicitly permits the operation of delivery robots, forcing tech and logistics companies to negotiate exemptions for pilot schemes. The EU's regulatory landscape is characterized by "soft law," where flexible guidelines are informed by ongoing dialogues between companies and local or national authorities. This approach allows for small-scale, limited-area deployments, often on private campuses, but it also highlights the absence of harmonized rules for the road and pavement use by delivery robots across Europe.

Given the region's interest in technological innovation and smart city initiatives, it's likely that individual countries are exploring regulatory frameworks to accommodate such technologies. The adoption and regulatory approach would vary by country, reflecting each nation's priorities in terms of innovation, public safety, and urban mobility solutions.

In the USA, regulation of autonomous delivery robots varies significantly by state, with at least 20 states having passed explicit statutes governing their operation. These laws were developed often after consultation with companies like FedEx and Amazon and tend to offer a high degree of operational leeway. Despite this, the regulatory landscape in the USA is fragmented, with different states setting varying speed limits and operational guidelines. This state-specific approach leads to a diverse regulatory environment across the country, with some local municipalities expressing concerns over accessibility and safety on public sidewalks.