The document discusses the challenges and opportunities in urban mobility and autonomous robotics. It highlights the market potential for food delivery and urban transportation while examining the competitive landscape and technological requirements for autonomy. Notably, it contrasts the regulatory and operational differences between automated driving and robotics, emphasizing the complexities involved in deploying robotic solutions in urban environments.

1. Robotica mobile in ambito urbano
Deep Learning Italia & Automata

2. Automata
Francesco, Michelangelo, Marco
Former colleagues at YAPE, now good friends and
co-founders
What do we do?
● Tech advisory to startups around robotics and
autonomous driving
● Teaching and promotion of STEM through
Masters, Webinars, Meetup and Masterclasses
● Mentoring

3. About Me
La Sapienza / UC Berkeley
Yape
INSEAD
Francesco Ricciuti
ricciuti.fr@gmail.com
https://www.linkedin.com/in/f-ricciuti/
Born and raised in Rome
BSc in Engineering Sciences @ Tor Vergata
MSc in Control and Automation Engineer @ La Sapienza
Visiting Student Researcher @ UC Berkeley
Robotics Software Engineer @ Yape
MBA Candidate @ INSEAD

4. this was in 2015!
Source: Wired
Source: Fortune
Autonomy is hard!

5. Where we are today
Starship Robotics
Founded in 2014, 3M autonomous deliveries to date. For comparison, it takes Deliveroo less than 4 days to do 3M
deliveries
Cruise
Raised $15.1B to date. Recently started fully autonomous service, covering only 70% of San Francisco

6. If it’s so damn hard, why are people still investing money in it?
Source: TechCrunch
Source: Cruise

7. The market need: food delivery
Food delivery companies are BLEEDING MONEY
Source: Just Eat annual report 2021
…but the market opportunity is huge!
€ 36B:
Food Delivery revenues,
Europe
€ 1,5B:
Food Delivery revenues,
Italy

8. The market need: urban mobility
Our cities are built around cars: 55% of the space dedicated to transportation is used by cars
The cycle of automobile dependency
Source: “Urban Space Distribution and Sustainable Transport”
5.3 deaths per
100k inhabitants:
Road fatalities, Italy
Driving is dangerous

9. The regulatory framework: autonomous robotics
countries with ongoing
commercial operations
countries with ongoing
experimental trials
Sources: Reuters, El Pais, Mirror, Business Korea, CTV News

10. The competitive landscape: autonomous robotics

11. Two different approaches
Remote first Autonomous first
Solve logistics and win empty market,
then think of unit economics
Cutting-edge technology, very cheap
deliveries but longer go-to-market times
(and costs…)

12. Whoami
Skills: Roboticists full-time,
Tech advisor and proud
co-founding member of Automata.
Name: Michelangelo Setaro
Contact: LinkedIn
Experience: Senior AD Software
Engineer at Stellantis and
former Lead Robotics Software
Engineer at Yape

13. What does autonomy means?
Let’s borrow some notions from automotive
We are here

14. High Level Design
Perception
Business Logic
Domain Specific
Communication
Diangostics
Watchdogs
Motion Planning
HAL
HMI
Actuation
Sensing
Hardware
Portrait of Autonomy
Cloud
Middleware
Software
Safety
Network
Remote
Supervision &
Teleoperations
Mapping &
Localization

15. System Design - Sensing Hardware
Nuro
● Lidar
● Ultrasonic
● Long & short range radar
● 360° Camera
Camello
● Ultrasonic
● Lidar
● Front Camera
Starship
● Ultrasonic
● Radar
● 12 Camera

16. Perception
Autonomy means perceiving and being aware of the sorroundings

17. Perception - Panoptic segmentation example
Everybody loves to watch a good video
Panoptic segmentation: DNN allow for pixel
level object information, thus fusing
together instance & semantic segmentation
Instance segmentation: identifies individual
objects within given category
Semantic segmentation: multiple objects of
the same category are considered one entity

18. Mapping and Localization
Autonomy means being aware of current position in space
Requires complex infrastructure and data
storage (1km x 1km map means at least 20GB)
Requires continuous data processing to
create/maintain maps
Higher the desired level of autonomy, more
complex maps become
Don’t trust who says “Tesla doesn’t use maps”
Google Cartographer - Pose Graph Slam

19. Mapping and Localization - Google Cartographer
Everybody loves to watch a good video cont’d
A great example of pose-graph slam, made by
google and Magazino
Works both with 3D and 2D lidar
Has a great cloud implementation that can be
a starting point for a centralized mapping
and localization infrastructure

20. Lacks of standard regulations, thus
difficulty to generalize and scale-up
Why autonomous driving is getting closer but not autonomous robotics?
Robotics
Automated driving is regulated by crystal
clear rules
Automotive
Petabytes of consolidated and constantly
updated SD/HD maps
Thousands of driving environment reference
datasets for training/testing/benchmarking
Each startup must rely on its proprietary
technology and infrastructure
Operating environment for urban robots is not
well defined, thus no dataset is really
general
Robotics is a capital intensive industry
with very demanding technological
constraints, few successful robotics
startup use the same approach of automotive
Main Challenges
Robotics vs Automotive