Mobile world congress in 2019. was just a slight improvement against the last year MWC. The main theme changed from selfdriving onto robotics and computer vision. Both selfdriving and robotics are powered by AI (Artificial intelligence). And 5G was everywhere (and real 5G modems,  not power point modems, as per Huawei).

Software defined radio

FaceBook, together with many operators is trying to support open source community to develop ecosystem that could compete against Huawei, named Telecom infrastructure project (TIP). One part of the TIP is CrowdCell which was presented in FaceBook’s booth, and for the purpose of demo, integrated with Athonet nVidia GPUs server.

CrowdCell is an SDR platform running on Lime mycrosystems SDR platform. The main value Lime claims to bring are their RF chips (as opposed to an usual SDR solution which would use a combination of different chips, which might not satisfy commercial power/ dimensions requirement, Lime packaged all RF baseband processing into a single chip). Lime investors include DFJ Capital of Steve Jurvetson fame. The suspicion is that VC is targeting AI + SDR combination. Myriad RF and bladeRF are using Lime chip, in combination with Altera FPGA. (Altera can be replaced by Xilinx if needed.)

Benetel is another SDR company, member of TIP, which beside base stations is doing UEs based on SDR. (Note: I was told about Benetel by Euricom.)

Artificial intelligence

Huawei

• 

• 

Huawei came out with full spectrum of AI chips (embedded) and cards (PCIe). (Note: For a brief moment I was there, it was almost an empty stand with more interest from Chinese onlookers than anyone else – presentation was in Chinese.)

The table below contains some comparison between Huawei AI and the rest.
(While Huawei effort is commendable, there’s a long way to catch up.)

Manufacturer	AI chip	TOPS	Power (Watt)
Huawei	Ascend 310 (SoC)	16TOPS Int8	8W
	Atlas 200 (?)	16TOPS Int8	9.5W
	Atlas 300 (PCIe)	64TOPS Int8	75W
nVidia	Xavier (SoC)	32TOPS (Int8?)	10W
	T4 (PCIe)	130TOPS Int8 260TOPS Int4	75W
Intel	EyeQ5 (SoC)	24TOPS (Int8?)	10W
Xilinx	Zync (RFSoC)	?	?
	Virtex VU9P (FPGA PCIe)	21TOPS Int8	75W
Google	TPU (?)	45TOPS Int8	?

Xilinx

Xilinx has on its stand an impressive Zynq RFSoC solution (8×8 MIMO) and Virtex xDNN FPGA card.

Main job for FPGA card is image recognition, and to lesser extent speech processing. There’s a new versal architecture on 7nm coming end of the year.
(There’s something strange with Xilinx numbers. It’s not clear why FPGA cards are so successful.)

ORBL

Orbl.io is a small Estonian company providing 5x DNNs to determine face, age, gender, emotions and liveliness of a person. Orbl sales are in Bay area.

Computer vision

Israeli stand

Israeli stand is again bigger this year. It always seems to be the most innovative and commercially ready stand.

Ionterra presented a SW solution for object recognition, tailored for low power devices (Raspberry pie or ARM, e.g.) which would achieve high FPS (Frames per second) without using DNNs. The nature of the algorithm was not revealed, but it’s clear to see the need (e.g. emergency shutdown of a robot has to be controlled locally.)

Elsight had a highly networked solution. E.g. 4x radio links of 4x different operators from a single base station.

Miscellanea

A Supermicro product manager was asked about DRAM pricing. The answer was that Supermicro has no pricing power when ordering memory and pays what suppliers ask. (If any negotiation takes place, memory supplier brings dubious arguments of fab flooding or similar, to maintain DRAM price.)
This DRAM contract price not going down is interesting. Definitely, the reduction in price is what would be expected based on DRAM price cycles.

One of Small cell forum vice-chairs is now with BEC technologies.

In the new technologies hall 8, an interesting information on a start-up doing quantum computing with entangled photons, based on silicon photonics. PsiQuantum (hiring) is led by ex Bristol University and Imperial college professors Jeremy O’Brien and Terry Rudolph (Erwin Schroedinger’s grandson).

ROS Essentials, Perception, and Control

Project: Search and sample return

 

Introduction to ROS

Packages and catkin workspaces

Biologically inspired robots

Write ROS nodes

Intro to kinematics

Forward and reverse kinematics

Denavit Hartenberg parameters for serial manipulators.

Project: Robotic arm: Pick and place

Human robot interaction & robot ethics

Perception overview

Introduction to 3D perception

Calibration, filtering and segmentation

1.  Intro to Calibration, Filtering, and Segmentation
2. Sensor Calibration
3. RGB Camera Model
4. Calibration Pattern
5. OpenCV Calibration
6. Extrinsic Calibration
7. RGBD Calibration in ROS
8. Point Cloud Filtering
9. Tabletop Segmentation Exercise
10.  Voxel Grid Downsampling
11. Pass Through Filtering
12. Segmentation in Perception
13. RANSAC Overview
14. RANSAC Plane Fitting
15. Extracting Indices
16. Outlier Removal FilterClustering and segmentation

Object recognition

Project: 3D perception

Soft robotics

Robot grasping

Introduction to controls

Quadrotor control using PID

1. Introduction to a Positional Controller
2. Quadrotor Kinematic and Dynamic Model 1
3. Quadrotor Quiz
4. Quadrotor Kinematic and Dynamic Model 2
5. Cascade PID Control
6. Lab Walkthrough
7. Environment Setup
8. Exploring the Sim
9. Helpful Tools
10. Completing PID Controller
11. Hover Controller
12. dynamic_reconfigure
13. Attitude Controller
14. Positional Controller
15. Lab Summary
16. PID Wrap Up

Swarm robotics

1. Introduction
2. READ : Research Papers
3. Swarm Robots in Medicine
4. Kilobots
5. Search and Rescue
6. Self-Assembly Swarm Robots
7. WATCH: Concepts in Action
8. Meet Sabine Hauert
9. Exclusive: Sabine Hauert
10. DO : Lab – Create a Swarm

Intro to neural networks

Deep neural networks

Convolutional neural networks

Fully convolutional networks

Lab: Semantic segmentation

Project: Follow me

Introduction to C++ for robotics

Those thinking 2018. MWC lacked sparkles, obviously didn’t see dazzling 3D Hypervsn(TM).

AI

All mobile processor manufacturers have added, next to graphical processors, Artificial intelligence (Neural networks) engine to their ARM processors.

• ARM added Mali
• MediaTek (in P60) added  APU
• Qualcomm added Adreno

There was no information on Huawei, and Apple AI processors – but they doubtlessly exist.

Israeli stand

An amazing inference chip, working entirely on optical principle was proposed by CogniFiber. (Each pixel has its own fiber of, say 30um, out of say, 600x400px. Thresholds of neural network are probably implemented as amplifiers for undersea fibers – no slowing down of light. Claim is: about 1000x less power and a few orders of magnitude speed up.)

Israeli syte.ai has a mobile app which can take camera photo to identify clothes (brand) on the person one envies and finds the price and where to buy it.

Another company has a ‘self-driving’ system for the passenger – one gets all the information about shops, restaurants, hotels, etc… on the camera photo, while being a co-driver.

SDR/URLLC

Fraunhofer institute demonstrated shortened TTI implemented in USRP.

CellXica from UK developed their own SDR platform and base station. (Thanx Prashant for tapping my shoulder.)

Other

SoftBank robotics brought childlike robots Nao and Pepper which were popular with women.

Memory

Intel True VR(TM) needs 2TB per hour (6 pairs of camera capture of 360degrees.)

Cars

Audi A8 has about 50x ARM processors.

(Ridiculous) concept car on Mercedes stand.

Graphene

Graphene corner was in Hall 8 just as the last year, and the monetisation remains the problem. Among other applications:

• Transparent graphene film over window’s surface. (Maybe in future we will have each window as a TV?)
• All kind of sensors
• Thermal conductor (better than copper)

Braincom-project.eu has increased the number of brain implanted electrodes from few hundred, to, using graphene, few tens of thousands of electrodes.

Russian stand

On quite modest Russian stand one could have seen Sailfish OS – alternative to iOS, Android and Windows – used by government. (It’s a Linux flavour.)

Path planning, concentrations and systems

Path planning

• Search:
  Discrete path planning and solving algorithms (A*).
• Prediction:
  Sensor fusion used to predict other objects behaviour.
• Trajectory generation in C++:
  Project: Path planning

Drive a car down a highway with other cars using one’s own path planner.

Advanced deep learning

• Fully connected convolutional networks
• Scene understanding
• Inference performance
• Project: Semantic segmentation

 

Autonomous vehicle architecture

• Introduction to ROS (Robot operating system):
  Architectural overview of ROS framework and setting up the environment.
• Packages & Catkin workspaces:
  ROS workspaces structure, essential command line tools and software package management
• Writing ROS nodes:
  Python and C++.
• Project: System integration project

Running the code on Carla, Udacity’s autonomous vehicle.

Inaugural class of Udacity Self-driving nanodegree graduation celebration with Sebastian Thrun.

Small Cell Forum and ITRI take leap forward in private networks

Technical support for the event was provided by SoftBank, ITRI and Magnum, a.i.

1. Meeting with Small ePC vendors, possible participants of Small cell forum Private ePC plugfest (in the role of Small cell forum Interoperability chair).
   

   ---

2. Status of DNN implementations
   (Note: companies maybe didn’t bring the best they have, otherwise, it’s bad):
   NXP had brought a range of radar chips, Lidar and ultrasonic chips. Sensor fusion is done in “blue box” powered by 4x ARM cores. (Implementation doesn’t seem perfect: Lidar doesn’t recognise all the people, and point cloud has artefacts, while camera recognises people where there are none.)
   Vodafone had Huawei‘s demonstrator for 5G (and possibly autonomous driving) running car simulation in PlayStation 4 (it seems Huawei doesn’t like x86 technology – I wouldn’t be surprised if Huawei and Sony announce partnership – for microprocessor/GPU/sensors, as Huawei is cut-off from top US tech.) Car simulation was not good/ convincing – if autonomous driving algorithms were used (and simulation had road-works/ cyclists/ pedestrians…).
   I observed a simulation similar to Huawei’s on Orange stand.
   HP Enterprise seemed to have car simulator similar to what we used, based on Unity engine. (Again, the driving skills of algorithm/ human were suboptimal.) HP Enterprise doesn’t seem to have a device for SMEs (a mistake, I think – they should be called HP for big Enterprise)
   Qualcomm had on their stand SnapDragon with DNN support through Adreno (GPU that Qcomm bought from AMD some time ago) manned by a Chinese person. A lonely stand, except for one more Chinese visitor – so the conversation was in Chinese. Ni hao. Not impressive (particularly in the light what other companies think about DNN on SnapDragon).
   SQREAM (from Israel) had SQL developed for GPUs. Very good solution. They mentioned (and the similar strategy will be used by all other companies) that they can swap AMD for Nvidia, but will never maintain two separate product lines (for AMD and Nvidia GPUs). Claim was that AMD improved a lot, but they stick with Nvidia. (And this is likely how AMD/ Nvidia “war” will end-up.)
   From Israeli stand, I got some indications how quick DNNs in mobile deployment, generally are.

   ---

3. SDR
   NI (National instruments) had a very nice 5G setup.
   LimeSDR, Octasic, …

   ---

4. I was introduced to James Tagg, the director of Penrose Institute in La Jolla (Sir Roger Penrose is the best physicist in the world, in my opinion). James is involved with two fascinating projects:

   To predict when NN can fail (not clear if this is a computable problem) – Anyone who trained NN knows it can fail unpredictably – solution to this problem will help DNN deployments tremendously.
   To prove professor Penrose’s objective reality of quantum physics, by gravity induced collapse of quantum wave functions through an experiment.
   How unlikely to meet such a person at an event like MWC. (Thank you James to introducing me to James, and thank you James for being so kind and having time for the chat.)

   ---

5. Other
   I spoke with Spanish satellite defence contractor, on the combination of Phased array antennae with Luneburg sphere.
   I met with Jean-Francois Lacasse of Cavium (who I first met through Small cell forum plugfest in Paris 2014) and had a nice chat – how things are turning great for multi-core ARM Cavium server. (He’s involved with LimeSDR and Ubuntu, and will chair Open Cellular)
   I visited 3D mapping companies (possibly to integrate their maps in Unity game engine). (LuxCarta ladies were especially nice, and gifted me with all sorts of gadgets.)

   In general, it seems more hype than money. Time for change…
   ( Hall 1: people who bought their way in.
   Hall 2: companies that GSMA seriously believe contribute to industry and are making money
   Hall 3: old companies, old money
   Hall 4: conference rooms
   Hall 5-7: small companies, and pavillions per country (Irish one is always the one with the best beer))

And when algorithms fail, below: