Abstract: A self-regulated solar power delivery system for data center. The ambient temperature outside of the data center is monitored. When the temperature exceeds a preset threshold, a controller activates switches to connect a PV system to a DC/DC converter and the DC/DC converter to a plurality of thermoelectric coolers (TECs). When the temperature drops below a second threshold, the controller disconnects the PV system. In this manner, when additional cooling is needed the most, i.e., during hot ambient temperature, the PV system also generates the most energy and can be used to energize TECs which enhance heat transportation from the processors. The PV system may also be used to activate a liquid cooling pump or other cooling devices to enhance heat removal from the servers.

Abstract: A two-phase fluid management system, may include a sealed container and a mobile condenser that moves within the sealed container. The sealed container may include a plurality of input ports, each to receive a two-phase fluid as vapor from a respective one of a plurality of IT enclosures and a plurality of output ports, each to return the two-phase fluid as liquid to the respective one of the plurality of IT enclosures. The mobile condenser may be coupled to or include an actuator to move the mobile condenser to a respective one of a plurality of positions within the sealed container. An air intake and air outlet of the mobile condenser may form a sealed connection with a pair of condenser ports when the mobile condenser moves to the respective one of the plurality of positions.

Abstract: The disclosure provides a system, for designing and developing immersion cooling in data centers. The system includes an internet technology (IT) tank that houses a computing device immersed in a two phase coolant; an aisle, in which the immersion tank is disposed, that captures a first vaporized portion of the two phase coolant escaped from the immersion tank; a condenser that transforms the first vaporized portion of the two phase coolant that escaped from the immersion tank into a first liquid portion of the two phase coolant; a second condenser captures and condenses a second portion of vapor; and a liquid distributor manages the cooling fluid and coolant fluid for the IT tank and two condensers.

Abstract: An adopting core device including a main board, a server connector module, a leaking sensor, and an electromagnet device is proposed in the current application. In an embodiment, a main board including a fluid channel assembled by a manual mating connector through hoses and a blind mating connector fixed on the other side. In an embodiment, the manual mating connector is connected to a rack connector of a rack manifold of an electronic rack coupled to an external cooling fluid source to receive and to return cooling fluid from and to the external cooling fluid source. For example, the blind mating connector is capable of being engaged with or disengaged from a server fluid connector of a server chassis. In an embodiment, the server chassis comprises a leaking sensor configured to detect leakage of the cooling fluid within the server chassis.

Abstract: According to one embodiment, a server chassis for immersion cooling includes a chassis frame having a chassis base and an adjustable server panel coupled to the chassis base to receive a server blade to be mounted on the adjustable server panel. The server blade includes a first portion and a second portion. The first portion has a first set of electronic components and the second portion has a second set of electronic components. The adjustable server panel is adjustably mounted along the chassis base, such that when the server chassis is deposited into a container having two-phase coolant therein, the first set of electronic components is at least partially submerged into a liquid region filled with the two-phase coolant, and the second set of electronic components is positioned within a vapor region above an immersion line defining the liquid region and the vapor region.

Abstract: Systems and methods are disclosed for data protection in a cluster of data processing accelerators (DPAs) using a policy that determines a static partition of resources in each DPA in the cluster communicatively coupled to a host device. Each DPA has sensitive (secure) and non-sensitive (non-secure) resources. The host device and a DPA can access all resources of the DPA. Other DPAs can only access non-sensitive resources of a DPA. The partition of resources within a DPA is static and may be implemented in hardware or firmware. Resources include memory, one or more processing modules such as key generators and cryptographic modules, caches, registers, and storage.

Abstract: In one embodiment, a set of LIDAR images are received representing the LIDAR point cloud data captured by a LIDAR device of an autonomous driving vehicle (ADV) at different point in times. Each of the LIDAR imagers is transformed or translated from a local coordinate system (e.g., LIDAR coordinate space) to a global coordinate system (e.g., GPS coordinate space) using a coordinate converter configured with a set of parameters. A first LIDAR reflection map is generated based on the transformed LIDAR images, for example, by merging the transformed LIDAR images together. The coordinate converter is optimized by adjusting one or more parameters of the coordinate converter based on the difference between the first LIDAR reflection map and a second LIDAR reflection map that serves as a reference LIDAR reflection map. The optimized coordinate converter can then be utilized to process LIDAR data for autonomous driving at real-time.

Abstract: Embodiments are disclosed of a battery management system (BMS). The BMS includes a pair of back-to-back transistors configured to selectively discharge battery cells of a battery pack; a high-side driver configured to control operation of the pair of back-to-back transistors; an analog front end (AFE) circuit configured to monitor a voltage of each cell within the battery pack; and a micro controller unit (MCU) configured to control and monitor the AFE circuit and provide over-voltage protection to the pair of back-to-back transistors.

Abstract: A method, apparatus, and system for filtering a point cloud generated by a LIDAR device in an autonomous vehicle is disclosed. A point cloud comprising a plurality of points is generated based on outputs of the LIDAR device. The point cloud is filtered to remove a first set of points in the plurality of points that correspond to noise based on one or more of: point intensity measurements, distances between points, or a combination thereof. Perception data is generated based on the filtered point cloud. Operations of the autonomous vehicle are controlled based on the perception data.

Abstract: Embodiments are disclosed of a rack cooling apparatus, more specially, a rack fluid control and management unit with a first pair of fluid lines including a first supply line and a first return line and a second pair of fluid lines including a second supply line and a second return line. The first supply line and the first return line are adapted to be coupled to a first rack manifold and the second supply line and the second return line are adapted to be coupled to a second rack manifold in the same rack as the first rack manifold. First and second bi-directional fluid lines fluidly couple the first supply line to the second supply line.

Abstract: A cooling unit to deliver liquid cooling to one or more peripheral devices is disclosed. The cooling unit includes one or more first mounting structures, the one or more first mounting structures are mounted onto the cooling unit, where the one or more first mounting structures are movable along a first direction. The cooling unit includes one or more cooling devices mounted to the one or more first mounting structures, where each of the one or more cooling devices is coupled to the one or more liquid distribution channels to receive cooling liquid to cool one or more peripheral devices of a server system. The cooling unit includes one or more liquid distribution channels to distribute cooling liquid within the server system, between the server system and other server systems, and/or between the server system and a liquid distribution system of an electronic rack.

Abstract: A cooling system can distribute fluid to and from IT equipment in an IT rack which can be understood as a server rack which houses multiple servers. The cooling system can include three distribution channels for the server rack. At least one of the distribution channels can be connected to valves that are arranged to operate that distribution channel in a bi-directional manner. This bi-directional distribution channel can switch operation if any of the other distribution channels becomes inoperative.

Abstract: An electronic rack includes condensing, coolant distribution, and server regions. The condensing region includes a condensing container housing condensing coils and a coolant container to contain two phase coolant. The coolant distribution region includes a set of rack manifolds having at least a rack liquid supply line to receive coolant from the coolant distribution region, and a vapor line to return vapor to the coolant distribution region, a liquid return line. The server region is coupled to the condensing region and the coolant distribution region, the server region includes a number of server slots to receive a number of servers, where each of the servers is at least partially submerged within two phase liquid coolant, where, when the servers operate, the servers generate heat that is extracted by the two phase liquid coolant thereby causing at least some of the two phase liquid coolant to turn into a vapor.

Abstract: Embodiments of the disclosure discloses a method and system for a virtualization environment for a data processing (DP) accelerator. In one embodiment, a data processing (DP) accelerator includes one or more statically partitioned resources and one or more virtual functions (VFs) each associated with one of the one or more statically partitioned resources. A virtual machine (VM) of a host is assigned one of the one or more VFs to access the statically partitioned resources associated with the assigned VF. The VM has no access to the rest of the one or more statically partitioned resources of the DP accelerator.

Abstract: A cooling unit includes a unit inlet and open-loop, and closed-loop outlets. A fluid bypass line is fluidly coupled to the unit inlet and to the open-loop and closed-loop outlets. A main cooling loop is fluidly coupled to the fluid bypass line includes a heat exchanger having an inlet and an outlet. The inlet of the heat exchanger is fluidly coupled by a supply line to the fluid bypass line, and the outlet of the heat exchanger is coupled by a return line to the fluid bypass line. A pump is fluidly coupled in the return line and a two-way line is fluidly coupled to the bypass line and to the return line upstream of the pump. Different system loop designs and operation modes are configured by adjusting three way valves and bidirectional loops.

Abstract: A rack with two-phase loop recirculation includes a supply manifold, a return manifold, and a separator. For example, a supply manifold is configured to receive two-phase cooling fluid from a cooling fluid source to distribute the two-phase cooling fluid to one or more server chassis. The two-phase cooling fluid is to extract heat from the one or more electronic devices and to transform into two-phase mixing fluid having at least a portion of the two-phase fluid transformed into vapor. A return manifold is configured to receive the two-phase mixing fluid from one or more loops associated with one or more electronic devices of the server chassis. A separator disposed on the return manifold is configured to separate the vapor of the two-phase mixing fluid and to divert first remaining two-phase cooling fluid of the two-phase mixing fluid directly back to the supply manifold through a return loop.

Abstract: According to one embodiment, a hybrid cooling system includes a cold plate that is arranged to mount on an IT component that is mounted on a piece of IT equipment, the cold plate is arranged to receive coolant via a supply line and to return warmed coolant via a return line, the warmed coolant is produced by the cold plate when the cold plate is in contact with the IT component and heat generated by the IT component is transferred into the coolant by the cold plate; a TEC element that is arranged to couple to the IT component; and a heat sink that includes a base that is arranged to couple to the TEC element and one or more fins, the TEC element is configured to transfer at least a portion of the heat generated by the IT component into the one or more fins of the heat sink.

Abstract: The cooling system for a data center includes an information technology (IT) container, secondary condensing system, and a coolant distribution unit. In particular, an IT container includes a liquid region to store cooling liquid, a vapor region to receive vapor evaporated from the cooling liquid, and a primary condenser disposed within the vapor region to condense the vapor. For example, the external cooling air or cooling liquid is controlled to be delivered to condensers. Further, a secondary condenser is coupled to the IT container via a vapor line to receive at least a portion of the vapor from the vapor region of the IT container and to condense the portion of the vapor. Furthermore, a coolant distribution unit is coupled to the IT container and the secondary condenser to store and to distribute the cooling liquid to the IT container.

Abstract: A method of navigating an autonomous driving vehicle (ADV) includes determining a target function for an open space model based on one or more obstacles and map information within a proximity of the ADV, then iteratively performing first and second quadratic programming (QP) optimizations on the target function. Then, generating a second trajectory based on results of the first and second QP optimizations to control the ADV autonomously using the second trajectory. The first QP optimization is based on fixing a first set of variables of the target function. The second QP optimization is based on maximizing a sum of the distances from the ADV to each of the obstacles over a plurality of points of the first trajectory, and minimizing a difference between a target end-state of the ADV and a determined final state of the ADV using the first trajectory.

Abstract: Learning disentangled representations is an important topic in machine learning for a wide range of applications. Disentangled latent variables represent interpretable semantic information and reflect separate factors of variation in data. Although generative models may learn latent representations and generate data samples as well, existing models may ignore the structural information among latent representations. Described in the present disclosure are embodiments to learn disentangled latent structural representations from data using decomposable variational auto-encoders, which simultaneously learn component representations and encode component relationships. Embodiments of a novel structural prior for latent representations are disclosed to capture interactions among different data components. Embodiments are applied to data segmentation and latent relation discovery among different data components. Experiments on several datasets demonstrate the utility of the present model embodiments.