Talking Tech with Rajkumar Jain, Vice President, Software Engineering, Maps, TomTom

How has the mapmaking process evolved over the years and what are the newest technologies used in it?

Customer needs change every day and mapmaking has evolved accordingly over the last decade. There was the batch way of mapmaking, and now it’s continuous process transactional mapmaking that supports continuous releasable maps. We are moving from local sources to global sources. From a few thousand to billions of observations processed per day. Customers are increasingly moving from uncompiled maps (offline) to online maps (using APIs), from full maps to incremental maps for delivery and compilation.

Eventually, we’ll be heading in the direction of software platforms for mapmaking (location platform systems). These can be run in the Cloud or on the edge to support self-healing maps that are updated automatically, based on AI and deep learning techniques. And these will be needed to support new business models such as mapping as a service and marketplace to trade the data.

Another thing to watch out for is companies' strategy on the adoption of open-source mapping platforms alongside their own mapmaking platform to enable a broader ecosystem, faster time to market for new use cases (private layers and editing partners) and increase crowdsourcing in map content creation (infinite elasticity).

Mapmaking is an art and different from classic product building. While building a map for a given geographic region, maintaining it is challenging because getting real-world signals is dependent on many sources. Aggregating signals from the real world and making a highly confident update in the map can only be scaled and sustained through machines. Technology is vital to meet the customer’s requirements of real-time map data, freshness, and continuously releasable maps.

In terms of technology, machine learning (supervized and unsupervized learning) and AI are significant as they help process mountains of data, and quickly detect changes and move the decisions to machines. The human role comes in non-deterministic situations: building high-quality training data set covering geographical variation. Data analytics completes the feedback loop for AI/ML algorithms. Other technology used in mapmaking include big data processing (Spark/Hadoop/Hive), cloud computing, micro services/nano services, event-driven architecture (e.g., increased use of Kafka) and data analytics. Other technologies would be DevOps, MLOps, Semantic modeling of data (ontology engineering), Domain Driven Design (DDD), Domain Specific Language (DSL), vector and tile-based map formats to reduce the file sizes and binary data formats like Avro, Protocol buffers, etc., and localization and positioning based on SLAM (Simultaneous Localization And Mapping) for 3D maps.

Technology needs to power the transactional mapmaking ecosystem to ensure updates at faster speeds, with higher map quality at a lower cost (per modification). In turn, this enables the continuous detection of changes, conducting updates, quality checks, and delivery of the freshest, most accurate content that users demand. This is done using data-driven decision-making by relying on end user metrics.

Mapmaking of the future will add other niche concepts like Reinforcement Learning (RL), where self-learning agents will code the map continuously, and quantum computing.

Can you throw some light on the different kinds of maps out there and what are their individual benefits?

Maps can be classified based on topographic and general reference maps (visuals/displays), as well as thematic maps (e.g., weather maps). Map visualization, depicting country boundaries (administrative regions), displays details closer to reality, supported by such features as land use, land cover, buildings in 2D, 2.5D and 3D, and 3D landmarks and water.

Maps could be categorized based on their usage. There are Standard Maps (SD maps), which are meant for human consumption with a reality-to-map error threshold of 1 to 20 meters. High-Definition Maps (HD maps) are meant for machine consumption with a reality-to-map error threshold of under 15cm. It is used for highly autonomous driving.

Specialized voice maps are possible too. They use text-to-speech technology, delivering navigation instructions clearly spoken in local dialects, helping the drivers keep their eyes safely on the road. Voice maps also enable Virtual Personal Assistants (VPAs).

Point of Interest (POI) Search maps help people find popular food, lodging and retail establishments, gas stations and electric vehicle charging points, and other convenience points.

Maps can be 2D (latitude and longitude), 2.5D (latitude, longitude and elevation for certain features) or 3D (latitude, longitude and elevation). They could vary in supporting automotive use cases based on SAE levels (0 to 5), which includes SD, ADAS (Advanced Driver Assisted Systems) and HD maps:

• SD Maps match the real world, providing freshness, richness, and quality.

• ADAS Maps provide in-vehicle systems with high-quality road information, including gradient, lanes, curvature and speed limits.

• HD Maps provide precise 3D visuals for safe, autonomous driving. For automated driving to be safe, HD maps are constantly updated in near real-time. With driving gradually shifting from humans to machines, the role and scope of digital maps must go beyond turn-by-turn navigation and be built specifically for self-driving vehicles.

Is TomTom a good alternative for Google Maps? If yes, then how?

At TomTom, we’re mapmakers providing location technology to drivers, carmakers, enterprises and developers. As an independent global company, we work with partners and companies across sectors. Our maps, navigation software, real-time traffic information and APIs enable smart mobility on a global scale.

Today, after 30 years of experience, we know the driver, have a navigation engine, are the leader in traffic information, respect privacy, and use customers’ data only to improve our products and services.

At the heart of our mapmaking is quality-controlled technology. TomTom’s unique transactional mapmaking platform allows real-time quality assurance and continuously releasable maps.

Our digital map data goes beyond the range of sensors, allowing driver assistance systems to better anticipate the road ahead. We are a market leader in ADAS map technology, powering more than three million SAE Level 1 and Level 2 automated vehicles. We provide in-vehicle systems with high-quality road information, including gradient, lanes, curvature, and speed limits.

TomTom solutions are scalable, which means our customers don’t need to buy all products. They can mix and match with our maps, or our navigation software, or one of our services to make smarter mobility decisions.

Here’s an overview of TomTom’s differentiators:

• Data privacy: We don’t monetize our customers’ data; it is only used to make the user experience better.

• Focus: Our focus is on creating a safer, cleaner, congestion-free world.

• Experience: We constantly innovate to keep delivering on our vision, while market dynamics have changed.

• Consumer brand and insights: Having sold over 100m navigation devices, we know our customers intimately, enabling us to anticipate and adapt to customer needs.

• Technology leadership: We create proprietary and innovative technology for maps, navigation and services such as traffic information.

• Independent: This enables us to work with all and any partners across both the automotive and the technology industries.

How will India’s new geospatial policy boost innovation?

The policy reform will have a huge impact on the country’s geospatial economy and in advancing digital India. The ability to collect data in the real-world becomes easier and the ease of accessibility to data (imagery, LiDAR, etc.) presents an enormous opportunity for industry and academia to solve pending problems. These could be in the area of last mile connectivity or evolving unstructured addresses in India to structured ones, amongst others.

Similarly, it opens up avenues for innovation, by bringing together public and private players in a partnership model. They can create solutions by leveraging the digitization of maps using specialized location technology. It will certainly fast-track innovative solutions for automotive, electric vehicles, retail, healthcare, telecom, agriculture, logistics and transportation.

How is TomTom’s Pune center contributing to TomTom mapmaking?

Mapmaking is a complex, intensive, and collaborative process. TomTom’s Pune center is responsible for SD mapmaking. Our sourcing, production and engineering teams are co-located to provide the best user experience for Navigation (passenger cars, fleets and logistics for trucks) and Search (POIs, Addressing, Geocoding etc.). We also enable active input from map communities using the Map Share Reporter platform and APIs. Our architects, data scientists, software and GIS engineers, product owners and cartographers innovate together, developing highly automated map-making processes. Machine learning, artificial intelligence, cloud computing and big data stacks are some of the many technologies that we use every day.

Tell us about your Pune facility?

Founded in 1991, in Amsterdam, TomTom invented an easy-to-use navigation device as a start-up. We’ve grown since into a leading global technology company while retaining our entrepreneurial spirit across our worldwide operations.

We’ve been in India for thirteen years and our Pune centre is one of the largest worldwide, with engineering, marketing, sourcing, production, sales, customer care and all shared services departments.

Our teams process petabytes of data with complex algorithms to optimize our maps and solve global mobility challenges. From building strong engineering teams and creating customer market proximity, we are proud of all our accomplishments, especially being one of the top 25* employers in the Indian IT sector.

*India’s Top 25 Best Workplaces in IT & IT-BPM 2020 by Great Place to Work ® Institute