Trade and Connectivity Challenge – 4th Edition

BACKGROUND

Labels need to be pasted on cosmetic products following the rules and regulations of the countries they are being exported to. This is a laborious and repetitive task, especially for large volumes of products.

Currently, factory workers need to manually un-box, place the label onto the product, and put the product back in the box and close it - which is time-consuming and tiring, so productivity decreases as time goes by.

By automating this process, it would be possible to ensure that the quality of the labelling is consistent, while ensuring that large volumes can be managed with the same level of productivity throughout the day.

---

REQUIREMENTS

The Robotics & Automation solution should address the below statements:

1. Boxes of different shapes and sizes can be placed into the solution
2. The flaps of the boxes should not interfere during picking and placing of products from and back into the boxes respectively
3. The solution should be capable of labelling the product at different positions based on product requirements (X, Y, Z datum, vertical or horizontal placement of labels etc.)
4. The solution should store multiple recipes/functions so that change over time between different products is minimized
5. The solution needs to capture cycle time and count, and extract it in flat file format for analysis purposes

---

DESIRED OUTCOME

A solution that can efficiently, accurately and cost-effectively execute the work autonomously with minimum human intervention.

BACKGROUND

AI-based applications currently provide real-time assistance to land vehicle drivers with a number of tasks, such as selecting optimal routes, and being aware of available roads and traffic conditions.

However, similar applications do not yet exist for vessels. Such applications would be very useful, especially in congested ports, to ensure no fuel is being wasted due to inefficiencies.

High speed internet and reporting on vessel positions are gradually becoming available in many ports, such as Singapore, which would make the development of AI-based applications for vessels feasible. The application should be user-friendly and preferably accessible on mobile devices, to ensure that mariners can use them easily and readily. It should also increase navigation safety.

---

REQUIREMENTS

• Available on Android and iOS mobile applications, as well as web browsers
• Able to receive a variety of data relating to navigation at sea, including nautical charts, vessel positioning and movement, as well as meteorological data etc.
• Able to collect relevant data for machine learning, so as to improve suggested routes and user-experience
• Approval from the relevant local and/or international authorities, to ensure that safety at sea is not compromised
• Application should be reasonably priced to ensure large volume of subscribers

---

DESIRED OUTCOME

• A mobile application with a user-friendly interface which provides real-time guidance to mariners, from one point to a static or moving destination (e.g., another vessel which may be manoeuvring to other locations).
• The route guidance should take into consideration all pertinent factors, including safety of navigation, weather and other real-time factors, such as movement of other vessels within the vicinity.
• Analytical reporting on the amount of energy saved or greenhouse gas reduction based on navigational data

BACKGROUND

Many of the data gathering and information processing activities needed to make trade decisions today are highly manual in the maritime industry. Not all data is available publicly, and the team has to rely on their networks to get the desired information/data, which may not be entirely accurate/updated. At the same time, the recommendations are usually subjective and depend on different members’ assessment of the current situation.

The ”ideal state” would be to have an automated decision recommendation engine, that takes into account past relationships across the different variables and expected future trends/trajectories to predict trade activity for a specific commodity.

---

REQUIREMENTS

It is currently difficult and time-consuming for the commercial teams to obtain the information needed to decide on where and how to deploy our vessels. This is because the decision is dependent on many factors – including, but not limited to, demand and supply situation, as well as the current positioning of vessels and committed contracts. The information is obtained through a mix of both public and informal data sources.

The assessments by individual members also tend to be highly subjective (50% experience, 50% gut feel), and there is no way to consistently replicate the decision-making process. This also results in potential loss of revenue as sub-optimal decisions may be made. The solution must include automation of the data gathering process, machine learning models that predict future trade activities for a specific commodity and a recommendation engine that suggests the various options for vessel deployment.

The minimum viable product (MVP) should be able to:

• Provide a platform that provides options to maximise the revenues based on forecasted demand-supply imbalances in the next 60-120 days for major locations and commodities, and freight rates trends.
• Demonstrate the automated data gathering process for publicly available data.

---

DESIRED OUTCOME

This will be a collaboration between IMC and the startup and primarily for use in-house. We will provide guidance to the selected startup on the factors and different scenarios to consider, as well as the data needed to make the decisions to build the machine learning model. Hence, the Intellectual Property (IP) of the model will belong to IMC. If the paid pilot is successful, we do not preclude the possibility of setting up a joint venture (JV) with the startup to commercialise the product to other companies.

BACKGROUND

To ensure we can achieve a net zero future, consumers can contribute by purchasing products which are manufactured and transported sustainably. While there currently exist labels and certificates that define how sustainable a product is, they often only account for the embodied carbon and not the entire product life cycle.

By providing end consumers with life cycle carbon footprint information, they can make a more informed decision in their purchase, therefore driving the transition to a greener society and supporting the decarbonization of the transportation industry - which currently accounts for 16% of global greenhouse gas emissions.

---

REQUIREMENTS

• Data range: from major transportation modes (primarily sea, land) and logistic sites
• Data quality: based on primary data sources (minimize use of default industry average data
• Data granularity: able to reach container shipment level or even SKU level
• Data reporting: able to provide comprehensive data visualization for analysis and assurance purposes
• Data methodology: able to comply with global standards of end-to-end supply chain emissions calculation

---

DESIRED OUTCOME

A digital solution capable of tracking and measuring emissions across end-to-end supply chain and reporting for assurance purposes.

BACKGROUND

The current checking process is manual and labour-intensive, thereby also resulting in possible human errors. Product attributes and packaging vary at an SKU level, requiring an intelligent system to identify and capture information across a broad range of products. Products being received need to be compared with orders placed to validate that the right products have been received.

---

REQUIREMENTS

1. Capture image of the product from various dimensions
2. Visual technology to read details on packaging and capture the specific attribute information, the position of which may vary across products
3. Ability to add new attributes to be checked as and when required
4. Product weighing
5. Integration with our in-house ERP system

---

DESIRED OUTCOME

A technology that can enable fast, accurate and seamless data capturing thereby reducing human reliance and error, as well as faster turnaround of goods.

A proposed automation/solution would look like/be able to:

1. Pick the product from the carton or from the infeed conveyor line
2. Scan the bar code and compare if it is part of the consignment document being inspected
3. Compare the image of the product with a master image and confirm the package version
4. Read the batch number, COO, ingredients, etc., and update the same into the Warehouse Management System (WMS)
5. If all the attributes are matching, then drop the product in the “accept” line, otherwise drop the product in the “reject” line for human intervention

BACKGROUND

Teleoperation is the use of human operators, manning a station at a command and control center, remote operating a piece of machinery or vehicle from afar. The remote operator has full control over the vehicle, handling the different driving mechanisms like decision making, steering, accelerating, braking and navigating.

To move goods within each warehouse space in our multi-site building at Toll City, industrial trucks (Forklifts, Reach Trucks) are commonly utilised. Each site has its own dedicated fleet of trucks. Like most resources, there are on peak and off peak periods for their utilisation, but often falling on the latter scenario. Therefore, we would like to consider optimizing the utilisation rate of these trucks, which could result in less trucks required overall, by sharing, managing and remote controlling them from a central location.

Most ready to implement teleoperating solutions in the market are retrofitted on autonomous vehicles/forklifts (AV), given the ease and availability of software integration methods (APIs) that are tied to specific control types. The use case is focused mainly to manage exceptions in case of AV breakdowns.

In our case, we like to apply to manually driven industrial branded forklifts only. This meant extra considerations have to be made in terms of installing the right fitted actuators, sensors and cameras to control existing driving mechanism in the manual forklift.

---

REQUIREMENTS

Technological / Performance criteria:

• During POC phase, this machine shall be placed and operated in a controlled/simulated environment where there will be no actual operation.
• Forklifts shall be able to perform the standard forklift mechanical actions e.g. lifting and stacking of pallets on level ground and onto a rack.
• The machine should be able to reach up to 2.5m/3m during the POC phase, in about the same time that human operator can achieve onsite.
• Forklifts should have a certain level of safety such as 360 view of the forklift, sensor for obstacle detection and prevention.
• The solution can be applied to a brand of choice by the start-up, but should demonstrate minimal effort to adapt to new forklift models.

Cost: TBD

Timeframe: 6-8 months

Potential market / business opportunity: All of Toll warehousing facilities could potentially apply this.

---

DESIRED OUTCOME

A teleoperating station sitting in innovation centre, that can oversee/control a forklift by a human to move, lift and stack goods:

• from point A to B on a flat ground
• from point A on a flat ground to a rack location
• from a rack location to point A on a flat ground

---

ADDITIONAL INFORMATION

Toll can provide:

• A test environment in our warehouse and network set up
• A working office in their innovation centre
• Developers experienced in mechanical and software design from the startup

BACKGROUND

In all warehouses, even with a Warehouse Management System (WMS) in place, the only way to confirm the inventory is to do a physical inventory count, by scanning or checking the digits with human operators during the cycle counting (such as with Radio Frequency (RF), voice and vision picking). Given the size of most warehouses, doing a physical inventory comes with many challenges. Manual inventories are extremely labour and time intensive, and as such are very expensive processes. Therefore, full wall to wall stock takes are only done periodically, whilst reliance on cycle counting for high value or fast moving items helps with inventory management on an ongoing basis.

The various processes of inventory management including stock takes and cycle counting, take considerable labour effort whilst also being subject to shortcomings in terms of accuracy or completeness.

Breakthroughs in computer vision AI in recent years have resulted in higher accuracy in object counting as a use case. However, when applied to visual inspection/counting of inventory goods in the warehouse, the use case is limited to pallet based counting and single deep racking type. In many warehouse operations within Singapore due to the distribution profile, handling and inventory are represented more by carton/case and split case consumption.

Therefore, we would like to extend the use case to provide a more accurate visual identification of inventory at the loose case/carton level, taking into consideration goods storage in double deep racking type as well as low lighting conditions within a warehouse.

---

REQUIREMENTS

• A computer vision AI detection model/solution to visually identify and count each case/carton object on every rack. However, additional sensors could also be considered as part of the solution to improve detection in a challenging environment.
• Counting accuracy should be of 95% or above
• The count can be recorded but is required to be automatically matched against WMS data at this stage
• Shelving type should include both single dip and double dip, also taking into consideration dark corners in the shelving unit under low lighting condition
• Hardware solution, integrated with required camera and sensor recording unit that can move along the aisle between rackings, as well as traverse the racks from bottom to top. Drone is an option but not limited to it.

Cost: TBD

Timeframe: 6-8 months

Potential market / business opportunity: All of Toll warehousing facilities could potentially apply this.

---

DESIRED OUTCOME

A working prototype product that is able to operate within one of our warehouse site fulfilling above requirements.

---

ADDITIONAL INFORMATION

Toll can provide:

• A test environment in our warehouse and network set up
• A working office in our innovation centre
• Developers experienced in mechanical and software design from the startup